US20160188578A9 - Method and apparatus for assembling a set of documents related to a triggering item - Google Patents
Method and apparatus for assembling a set of documents related to a triggering item Download PDFInfo
- Publication number
- US20160188578A9 US20160188578A9 US13/182,245 US201113182245A US2016188578A9 US 20160188578 A9 US20160188578 A9 US 20160188578A9 US 201113182245 A US201113182245 A US 201113182245A US 2016188578 A9 US2016188578 A9 US 2016188578A9
- Authority
- US
- United States
- Prior art keywords
- documents
- user
- electronic
- triggering item
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G06F17/30011—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/93—Document management systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/338—Presentation of query results
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9538—Presentation of query results
-
- G06F17/30696—
-
- G06F17/30705—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G06K9/00483—
-
- G06K9/6202—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/41—Analysis of document content
- G06V30/418—Document matching, e.g. of document images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/535—Tracking the activity of the user
Definitions
- the present invention relates generally to data management, and relates more particularly to technology for assisting in data management.
- Events such as business trips, employment interviews, proposal activities, technical paper reviews, and the like often require that an individual review a set of documents related to the event. For example, interviewing a job candidate may require review of the candidate's resume, transcripts, writing samples, and other documents.
- the present invention relates to a method and apparatus for assembling a set of documents related to a triggering item.
- One embodiment of a method for assembling a set of electronic documents related to an electronic triggering item detected by a computing device being operated by a user includes automatically extracting by the computing device a set of features from the triggering item, without receiving a request by the user to assemble the set of electronic documents, and assembling as the set of electronic documents a plurality of documents that is relevant to the set of features, wherein the plurality of documents is retrieved from a plurality of different types of electronic sources.
- FIG. 1 is a schematic diagram illustrating one embodiment of a system for assembling a set of documents related to a triggering item, according to the present invention
- FIG. 2 is a schematic diagram illustrating a second embodiment of a system for assembling a set of documents related to a triggering item, according to the present invention
- FIG. 3 is a flow diagram illustrating one embodiment of a method for assembling a set of documents related to a triggering item, according to the present invention.
- FIG. 4 is a high level block diagram of the present invention implemented using a general purpose computing device.
- the present invention relates to a method and apparatus for assembling a set of documents related to a triggering item.
- the triggering item may be, for example, an event or a document that is relevant to the user.
- embodiments of the invention harvest documents from a plurality of sources and then estimate the relevancy of the harvested documents to the triggering item. Once the relevancies are estimated, the harvested documents are ranked and/or filtered based on relevancy to produce the set of documents.
- the set of documents is then delivered to the user, who may provide feedback that helps guide the assembly of future document sets.
- a “document” refers to any sort of information that can be harvested from a computing device or network.
- a set of documents might include things like emails, white papers, and contracts as well as flight information, news articles, and weather data.
- FIG. 1 is a schematic diagram illustrating one embodiment of a system 100 for assembling a set of documents related to a triggering item, according to the present invention.
- the system 100 is implemented within a computing device that is operated by a user.
- the computing device may comprise, for example, a desk top computer, a lap top computer, a tablet computer, a server, a smart phone, a gaming device, a set top box, a digital media receiver, or the like.
- the computing device may or may not be connected to a network.
- the inputs to the system 100 comprise a set of documents and a triggering item.
- the system outputs a subset of the documents that is relevant to the triggering item.
- the system 100 comprises five main components: a harvester 102 , a full text query (FTQ) processor 104 , a classifier 106 , a set builder 108 , and a plurality of templates 110 .
- FTQ full text query
- the harvester 102 indexes and processes documents including files (e.g., word processing files, spreadsheet files, presentation files, individual slides in presentation files, audio files, video files, etc.), calendar events, to do lists, notes, emails, and email attachments. These documents may be retrieved locally from the user's computer and/or remotely from network storage (e.g., a server that stores documents produced by a plurality of users). In the latter case, the harvester 102 may also retrieve documents from the World Wide Web (e.g., web pages).
- the harvester 102 is coupled to the FTQ processor 104 and the classifier 106 , such that the harvester 102 may provide harvested documents to the FTQ processor 104 and the classifier 106 .
- the FTQ processor 104 is coupled to the harvester 102 .
- the FTQ processor also receives triggering items from the computing device.
- a triggering item may comprise an event (e.g., a calendar item, a business trip, an employment interview, a proposal activity, a technical paper review, etc.) or a document (e.g., an email message, a word processing document, a spreadsheet, a slideshow presentation, a text file, a web page, a chat message, etc.).
- the triggering item may be entered automatically by an application running on the computing device (e.g., a scheduling application).
- the FTQ processor 104 estimates the relevancy of the triggering item to the harvested documents.
- the FTQ processor extracts features from the triggering item (such as keywords, people, dates, locations, acronyms, or the like) and then builds a query around these features.
- the features are assigned weights in the query.
- the query is then run against a search (e.g., a LUCENE full text query) over the harvested documents. This produces a list of ranked documents based upon the features (words) and document frequency.
- the FTQ processor 104 is coupled to the set builder 108 and provides this ranked list to the set builder 108 .
- the classifier 106 also receives the triggering item.
- the classifier 106 uses information about the triggering item to select a template from the templates 110 .
- the classifier 106 extracts an activity type from the triggering item, and then selects the template that most closely matches the activity type.
- the activity type guides the identification of relevant documents. That is, for certain events and documents (e.g., business trips), the same types of documents must typically be gathered (e.g., flight details, hotel and car reservations, weather forecast, itinerary, etc.).
- the template identifies these types of documents.
- the classifier also receives the harvested documents.
- the classifier 106 extracts an activity type from each harvested document, much in the same way that the classifier 106 extracts the activity type from the triggering item.
- the classifier 106 is a maximum entropy classifier.
- the classifier 106 is coupled to the set builder 108 and provides the selected template and the document activity types to the set builder.
- the set builder 108 is coupled to the FTQ processor 104 and to the classifier 106 and receives the list of ranked documents and the selected template from the FTQ processor 104 and the classifier 106 , respectively.
- the set builder 108 uses the. selected template and the activity types to re-rank and filter the list of ranked documents. In particular, the set builder 108 re-ranks the documents based on those whose activity type most closely matches the activity type of the selected template.
- the set builder 108 outputs a final set of documents relevant to the triggering item.
- the set builder 108 receives user feedback regarding the final set of documents. The user feedback may be used to refine the assembly of future sets of documents, as discussed in further detail below.
- FIG. 1 illustrates only one possible configuration of the system 100 .
- two or more of the individual system components could be replaced by a single component.
- any single one of the system components could be replaced by multiple components.
- FIG. 2 is a schematic diagram illustrating a second embodiment of a system 200 for assembling a set of documents related to a triggering item, according to the present invention.
- the system 200 is implemented within a computing device that is operated by a user and that may or may not be connected to a network.
- the inputs to the system 200 comprise a set of documents and a triggering item.
- the system outputs a subset of the documents that is relevant to the triggering item.
- the system 200 comprises four main components: a harvester 202 , a full text query (FTQ) processor 204 , a pipeline 206 comprising a plurality of processors 210 1 - 210 n (hereinafter collectively referred to as “processors 210 ”), and a set builder 208 .
- the harvester 202 indexes and processes documents including files, calendar events, to do lists, notes, emails, and email attachments. These documents may be retrieved locally from the user's computer and/or remotely from network storage. In the latter case, the harvester 202 may also retrieve documents from the World Wide Web (e.g., web pages).
- the harvester 202 is coupled to the FTQ processor 204 and the pipeline 206 , such that the harvester 202 may provide harvested documents to the FTQ processor 204 and the pipeline 206 . ( 0026 )
- the FTQ processor 204 is coupled to the harvester 202 .
- the FTQ processor also receives triggering items from the computing device. As discussed above, a triggering item may comprise an event or a document. The triggering item may be entered directly by the user or may be entered automatically by an application running on the computing device.
- the FTQ processor 204 estimates the relevancy of the triggering item to the harvested documents.
- the FTQ processor extracts features from the triggering item (such as keywords, people, dates, locations, acronyms, or the like) and then builds a query around these features.
- the features are assigned weights in the query.
- the query is then run against a search (e.g., a LUCENE full text query) over the harvested documents. This produces a set of raw data that can be used to relate the harvested documents to the triggering item.
- the FTQ processor may also rank the harvested documents based upon the features (words) and document frequency.
- the FTQ processor 204 is coupled to the pipeline 206 and provides the raw data (or the ranked list) to the pipeline 206 .
- the pipeline 206 also receives the triggering item.
- the pipeline 206 uses information about the triggering item and the raw data (or the ranked list of documents) produced by the FTQ processor 204 to filter and rank (or re-rank) the documents.
- the pipeline 206 comprises a plurality of pipelined processors 210 (i.e., the output of one processor 210 passes as input to the next processor in the series).
- each of the processors 210 contains its own persistent state (model, classifier, etc.).
- the processors 210 are arranged such that the first processor 210 1 establishes an initial set of scores or rankings for the documents, and the subsequent processors 210 2 - 210 n adjust these scores.
- Each of the processors 210 may adjust the scores based on a different algorithm or model. For example, the processors 210 adjust the scores based on text frequency/inverse document frequency, proximity, or object recency, among other criteria.
- the pipeline 206 is coupled to the set builder 208 and provides the list of filtered and re-ranked documents to the set builder.
- the set builder 208 is coupled to the pipeline 206 and receives the list of filtered and ranked documents from the pipeline 206 .
- the set builder 208 outputs a final set of documents relevant to the triggering item.
- the system 200 additionally comprises a feedback bus 212 that receives user feedback regarding the final set of documents.
- the feedback bus is coupled to at least one of the processors 210 in the pipeline 206 and to the set builder 208 .
- the set builder 208 or any coupled processor 210 may access the user feedback that is available via the feedback bus 212 .
- the user feedback may be used to refine the assembly of future sets of documents.
- the user feedback may be provided to any one or more of the processors 210 (e.g., automatically or responsive to a request from one or more of the processors 210 ).
- the processors 210 are capable of maintaining their own persistent states, the user feedback can be used to help train or adjust the algorithms or models used by the processors 210 .
- FIG. 2 illustrates only one possible configuration of the system 200 .
- two or more of the individual system components could be replaced by a single component.
- any single one of the system components could be replaced by multiple components.
- FIG. 3 is a flow diagram illustrating one embodiment of a method 300 for assembling a set of documents related to a triggering item, according to the present invention.
- the method 300 may be implemented, for example, by the systems 100 and 200 illustrated in FIGS. 1 and 2 . As such, reference is made in the discussion of FIG. 3 to various elements of FIGS. 1 and 2 . It will be appreciated, however, that the method 300 is not limited to execution within a system configured exactly as illustrated in FIG. 1 or FIG. 2 and, may, in fact, execute within systems having alternative configurations.
- the method 300 is initialized at step 302 and proceeds to step 304 , where the harvester 102 (or 202 ) harvests documents including files (e.g., word processing files, spreadsheet files, presentation files, individual slides in presentation files, etc.), calendar events, to do lists, notes, emails, and email attachments. These documents may be retrieved locally from the user's computer and/or remotely from network storage. In the latter case, the harvested documents may also include documents retrieved from the World Wide Web (e.g., web pages).
- files e.g., word processing files, spreadsheet files, presentation files, individual slides in presentation files, etc.
- calendar events e.g., calendar events, to do lists, notes, emails, and email attachments.
- the FTQ processor 104 and the classifier 106 receive a triggering item (e.g., a document or event).
- a triggering item e.g., a document or event.
- the system 100 (or 200 ) proactively detects the triggering item and assembles the documents (i.e., without receiving a request from the user to assemble the documents).
- the system 100 (or 200 ) may detect the addition of a new item to the user's calendar and automatically assemble a set of relevant documents.
- the triggering event comprises the user accessing an application that assembles the set of documents.
- the classifier 106 extracts an activity type from the triggering item and from each of the harvested documents.
- the activity type may be extracted from the harvested documents in advance of the receipt of the triggering item (i.e., the activity type is not necessarily extracted from the harvested documents simultaneously with the extraction of the activity type from the triggering item).
- harvesting and classification of documents is performed substantially continuously (e.g., on a long-term basis), even when there are no triggering items on which to operate.
- the activity type is determined by first parsing the text of the triggering item or the document into separate words (removing punctuation and stop words). An information or semantic extraction algorithm may be used to find word types.
- the classifier 106 determines the activity type by classifying the triggering item or the document in accordance with the parsed words. In one embodiment, the classification is performed in accordance with a maximum entropy classifier. In another embodiment (e.g., where the method 300 executes within the system 200 illustrated in FIG. 2 , one or more of the processors 210 may extract the activity type).
- the FTQ processor 104 (or 204 ) builds a query in accordance with features extracted from the triggering item.
- the features may be, but are not necessarily, the same as the words that are parsed for classification purposes.
- the extracted features include at least one of: a keyword, a person, a date, a location, or an acronym. Stop words are discarded.
- the query is built by assigning weights to the extracted features.
- the query is a Lucene full text query.
- step 312 the FTQ processor 104 (or 204 ) runs the query against a search over the harvested documents. This produces a list of ranked documents based upon the extracted features (words) and document frequency.
- the classifier 106 selects a template in accordance with the activity type of the triggering item.
- the set of templates 110 includes a plurality of templates associated with different activity types.
- a given template identifies, for the associated activity type, the kinds of documents that a user would typically wish to gather.
- a template for the activity type “business trip” might list the following types of documents: flight information, hotel reservation, car reservation, weather forecast, contact information, and the like.
- the system 100 is pre-trained with knowledge of certain activity types and their associated templates; however, the system 100 can also be trained to learn new activity types and their associated templates.
- the classifier 106 selects the appropriate template by using the triggering item's activity type as a key into a template hash table.
- the set builder 108 re-ranks and filters the list of ranked documents produced by the FTQ processor 104 , in accordance with the template selected by the classifier 106 and the activity types of the documents in the list of ranked documents. In one embodiment, the set builder 108 does this by comparing the activity type of each document to the activity type of the selected template. The ranks of the documents whose activity types match the activity type of the template are increased, while the ranks of the documents whose activity types do not match the activity type of the template are reduced (or the documents are removed from the list). In the case of the system 200 , the re-ranking and filtering is performed by the pipeline 206 , in accordance with the algorithms or models implemented in each of the processors 210 . In this instance, it is possible that no templates may be used.
- ranking the documents includes recency ranking, word popularity, and learning.
- An initial ranking may be based on a combination of recency ranking and word popularity.
- Recency ranking assigns a higher weight to documents having later modification dates, access dates, and/or number of accesses.
- Word popularity uses a global dictionary containing high frequency words and weights. The number of high frequency words is user-configurable and may be set to a finite number of words (e.g., fifty thousand). When a new document is harvested, keywords that do not exist in the dictionary are added to the dictionary.
- stemming is applied to obtain root words, and all words are converted to the lowercase letters.
- Weights in the dictionary may be updated continuously as the system 100 (or 200 ) operates. Keywords appearing in frequently accesses documents may be assigned higher weights, while keywords appearing in less frequently accessed documents may be assigned lower weights. In this way, the system 100 (or 200 ) has access to an up-to-date and accurate model of the user's universe and behavior.
- the re-ranking and filtering performed in step 316 produces a final set of documents.
- the final set of documents comprises at least one document that the system 100 (or 200 ) believes is relevant to the triggering item.
- the final set of documents may also include an alert to notify the user of potentially missing information (e.g., no hotel reservation was located for a business trip).
- the system 100 (or 200 ) could interface with a task learning application that determines the appropriate procedure to execute to retrieve the missing document.
- the final set of documents is output by the set builder 108 (or 208 ) in step 318 .
- the set builder 108 optionally receives user feedback regarding the final set of documents.
- the user feedback may be implicit or explicit.
- Explicit feedback might include, for example, a user command to keep a given document in the final set of documents or to remove a given document from the final set of documents.
- Implicit feedback might include, for example, the fact the a user did or did not open a given document in the final set of documents, that the user opened a folder containing a given document in the final set of documents, or asked for items that are similar to a given document in the final set of documents.
- the set builder 108 stores the user feedback. Storage of the user feedback involves adjusting the way in which the set builder 108 builds sets of documents.
- the set builder 108 updates the global dictionary (discussed above) at the word level. For example, the weights of the most significant words in the document corresponding to the user feedback may be increased if the user feedback is positive, or decreased if the user feedback is negative.
- keywords from the query built in step 310 may be added to the document as metadata. If the keywords also exist in the global dictionary, their weights may also be adjusted as described above.
- the global dictionary may then be used to re-rank sets of documents assembled in response to future queries.
- the metadata stored with the documents may also be used to increase the rank of the documents (e.g., when the metadata matches future query keywords, the documents' ranks are increased).
- the set builder 108 makes updates at the document level. For example, each document in the final set of documents may be associated with a signature of the triggering item, where the signature also indicates whether the user feedback with respect to the document was positive or negative. If the document then shows up in a future list of ranked documents, the set builder 108 (or 208 ) will compare the signature to the signature of the current triggering item. If the signatures match and the user feedback was positive, the document's ranking is increased; if the signatures match and the user feedback was negative, the document's ranking is decreased. In one embodiment, explicit feedback has a more significant effect on ranking than implicit feedback.
- the set builder 208 may provide the user feedback to one or more of the processors 210 .
- any of the processors 210 may adjust the algorithms or models implemented therein in accordance with the user feedback.
- the method 300 terminates in step 324 .
- the system 100 may be pre-trained with knowledge of certain activity types and their associated templates. However, the system 100 can also be trained by the user to learn other activity types and their associated templates. To do this, the user identifies a new document and its associated activity type. This information is fed to the classifier 106 , which subsequently parses the new document for words and word types (e.g., dates, times, names, etc.). An information or semantic extraction algorithm may be used. The activity type, words, and word types are then stored by the classifier 106 as a new class.
- words and word types e.g., dates, times, names, etc.
- An information or semantic extraction algorithm may be used.
- the activity type, words, and word types are then stored by the classifier 106 as a new class.
- the system 100 can also be trained to recognize new documents of a known activity type.
- a new document is parsed into words and word types as described above. These words and word types are used, along with the known activity type, as training data for the classifier 106 .
- the systems 100 and 200 and associated method 300 may be deployed in substantially any computing device or network.
- the system 100 or 200 may be deployed in a home or office network.
- the system 100 or 200 can interact with other “smart” devices.
- a “smart” refrigerator that is connected to a home network may be able to detect what grocery items are in short supply, or a printer may be able to detect when it is about to run out of ink.
- the user may query the system 100 or 200 for a shopping list (e.g., where the triggering item is the event “grocery shopping”).
- the system 100 or 200 may further search for coupons for the items on the shopping list or for prices.
- the final set of documents presented to the user might include a shopping list, a set of coupons, and a list of stores advertising the lowest prices for the items on the shopping list.
- the system 100 or 200 may be configured to continuously update the user with information that is relevant to his interests.
- the triggering item may be the user's musical interests, the user's children's extracurricular activities, the user's upcoming vacation, the user's favorite sports teams, the user's current location, or the like.
- the user does not have to search for this information; rather, the information is automatically delivered by the system 100 or 200 , which proactively assembles a set of personalized documents relating to current information that is of interest to the user.
- Sorting and/or ranking of this information may be based on the time of day and/or the user's interest level (e.g., as expressed in a user profile). For instance, if the current time of day falls within the user's normal working hours, work-related information (e.g., “John just scheduled a meeting for 2:00 P.M. tomorrow”) might be ranked more highly than information that is not work related (e.g., “Tickets go on sale for the Giants game in ten minutes”).
- work-related information e.g., “John just scheduled a meeting for 2:00 P.M. tomorrow”
- the type of information provided, as well as the manner and time in which the information is provided may by learned dynamically based on user feedback or based on a user profile.
- the set of personalized documents is assembled on a periodic basis (e.g., every thirty minutes).
- the set of personalized documents is assembled as new relevant information is detected.
- Time sensitive documents may be removed from the set as they become outdated, so that the storage required by the system 100 or 200 is relatively low.
- this service is activated only when the user accesses the system 100 or 200 .
- the set of personalized documents could be stored (e.g., on a web site) for viewing at a time of the user's choosing.
- the set of personalized documents is provided as part of a notification service that notifies the user when new documents are available.
- the set of personalized documents may also include hyperlinks or other executable code that allows the user to share one or more documents in an email or on a social networking site (e.g., email a family member to let them know that a flight has been delayed), to add an event to the user's calendar (e.g., add an entry for a child's upcoming soccer game), to initiate a commercial transaction (e.g., purchase tickets to an upcoming concert), or to simply obtain more information about a given document (e.g., read the full news article about a particular event).
- a social networking site e.g., email a family member to let them know that a flight has been delayed
- an event e.g., add an entry for a child's upcoming soccer game
- to initiate a commercial transaction e.g., purchase tickets to an upcoming concert
- simply obtain more information about a given document e.g., read the full news article about a particular event.
- the set of personalized documents might resemble the following:
- this information may be grouped and/or ranked.
- the above set of personalized documents might resemble the following:
- FIG. 4 is a high level block diagram of the present invention implemented using a general purpose computing device 400 .
- a general purpose computing device 400 comprises a processor 402 , a memory 404 , a document assembly module 405 , and various input/output (I/O) devices 406 such as a display, a keyboard, a mouse, a modem, a microphone, speakers, a touch screen, and the like.
- I/O devices 406 such as a display, a keyboard, a mouse, a modem, a microphone, speakers, a touch screen, and the like.
- at least one I/O device is a storage device (e.g., a disk drive, an optical disk drive, a flash drive).
- embodiments of the present invention can be represented by one or more software applications (or even a combination of software and hardware, e.g., using Application Specific Integrated Circuits (ASIC)), where the software is loaded from a storage medium (e.g., I/O devices 406 ) and operated by the processor 402 in the memory 404 of the general purpose computing device 400 .
- a storage medium e.g., I/O devices 406
- a non-transitory computer readable medium e.g., RAM, magnetic or optical drive or diskette, and the like.
- one or more steps of the methods described herein may include a storing, displaying and/or outputting step as required for a particular application.
- any data, records, fields, and/or intermediate results discussed in the methods can be stored, displayed, and/or outputted to another device as required for a particular application.
- steps or blocks in the accompanying Figures that recite a determining operation or involve a decision do not necessarily require that both branches of the determining operation be practiced. In other words, one of the branches of the determining operation can be deemed as an optional step.
Abstract
Description
- This application claims the priority of U.S. Provisional Patent Application Ser. No. 61/415,722, filed Nov. 19, 2010, which is herein incorporated by reference in its entirety.
- This application was made with Government support under contract no. FA8750-07-D-0185 awarded by the Air Force Research Laboratory. The Government has certain rights in this invention.
- The present invention relates generally to data management, and relates more particularly to technology for assisting in data management.
- Events such as business trips, employment interviews, proposal activities, technical paper reviews, and the like often require that an individual review a set of documents related to the event. For example, interviewing a job candidate may require review of the candidate's resume, transcripts, writing samples, and other documents.
- These relevant documents are typically collected by performing a desktop search that requires several iterations. Such searches rely heavily on what the searcher knows and what he finds during the search. For example, the searcher may need to know where on his desk top the relevant documents reside (e.g., in a desk top folder, in an email message, on a networked server, etc.). Thus, such searches are time consuming and do not guarantee that the searcher will locate all relevant documents even after many iterations of searching.
- The present invention relates to a method and apparatus for assembling a set of documents related to a triggering item. One embodiment of a method for assembling a set of electronic documents related to an electronic triggering item detected by a computing device being operated by a user includes automatically extracting by the computing device a set of features from the triggering item, without receiving a request by the user to assemble the set of electronic documents, and assembling as the set of electronic documents a plurality of documents that is relevant to the set of features, wherein the plurality of documents is retrieved from a plurality of different types of electronic sources.
- The teachings of the present invention can be readily understood by considering the following detailed description in conjunction with the accompanying drawings, in which:
-
FIG. 1 is a schematic diagram illustrating one embodiment of a system for assembling a set of documents related to a triggering item, according to the present invention; -
FIG. 2 is a schematic diagram illustrating a second embodiment of a system for assembling a set of documents related to a triggering item, according to the present invention; -
FIG. 3 is a flow diagram illustrating one embodiment of a method for assembling a set of documents related to a triggering item, according to the present invention; and -
FIG. 4 is a high level block diagram of the present invention implemented using a general purpose computing device. - To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures.
- The present invention relates to a method and apparatus for assembling a set of documents related to a triggering item. The triggering item may be, for example, an event or a document that is relevant to the user. In particular, embodiments of the invention harvest documents from a plurality of sources and then estimate the relevancy of the harvested documents to the triggering item. Once the relevancies are estimated, the harvested documents are ranked and/or filtered based on relevancy to produce the set of documents. The set of documents is then delivered to the user, who may provide feedback that helps guide the assembly of future document sets. Within the context of the present invention, a “document” refers to any sort of information that can be harvested from a computing device or network. Thus, a set of documents might include things like emails, white papers, and contracts as well as flight information, news articles, and weather data.
-
FIG. 1 is a schematic diagram illustrating one embodiment of asystem 100 for assembling a set of documents related to a triggering item, according to the present invention. Thesystem 100 is implemented within a computing device that is operated by a user. The computing device may comprise, for example, a desk top computer, a lap top computer, a tablet computer, a server, a smart phone, a gaming device, a set top box, a digital media receiver, or the like. The computing device may or may not be connected to a network. - In general, the inputs to the
system 100 comprise a set of documents and a triggering item. In turn, the system outputs a subset of the documents that is relevant to the triggering item. As illustrated, thesystem 100 comprises five main components: aharvester 102, a full text query (FTQ)processor 104, aclassifier 106, aset builder 108, and a plurality oftemplates 110. - The harvester 102 indexes and processes documents including files (e.g., word processing files, spreadsheet files, presentation files, individual slides in presentation files, audio files, video files, etc.), calendar events, to do lists, notes, emails, and email attachments. These documents may be retrieved locally from the user's computer and/or remotely from network storage (e.g., a server that stores documents produced by a plurality of users). In the latter case, the
harvester 102 may also retrieve documents from the World Wide Web (e.g., web pages). Theharvester 102 is coupled to the FTQprocessor 104 and theclassifier 106, such that theharvester 102 may provide harvested documents to the FTQprocessor 104 and theclassifier 106. - The
FTQ processor 104, as discussed above, is coupled to theharvester 102. In addition to the harvested documents, the FTQ processor also receives triggering items from the computing device. As discussed above, a triggering item may comprise an event (e.g., a calendar item, a business trip, an employment interview, a proposal activity, a technical paper review, etc.) or a document (e.g., an email message, a word processing document, a spreadsheet, a slideshow presentation, a text file, a web page, a chat message, etc.). The triggering item may be entered automatically by an application running on the computing device (e.g., a scheduling application). - The FTQ
processor 104 estimates the relevancy of the triggering item to the harvested documents. In one embodiment, the FTQ processor extracts features from the triggering item (such as keywords, people, dates, locations, acronyms, or the like) and then builds a query around these features. In one embodiment, the features are assigned weights in the query. The query is then run against a search (e.g., a LUCENE full text query) over the harvested documents. This produces a list of ranked documents based upon the features (words) and document frequency. The FTQprocessor 104 is coupled to theset builder 108 and provides this ranked list to theset builder 108. - The
classifier 106 also receives the triggering item. Theclassifier 106 uses information about the triggering item to select a template from thetemplates 110. Specifically, theclassifier 106 extracts an activity type from the triggering item, and then selects the template that most closely matches the activity type. As discussed in further detail below, the activity type guides the identification of relevant documents. That is, for certain events and documents (e.g., business trips), the same types of documents must typically be gathered (e.g., flight details, hotel and car reservations, weather forecast, itinerary, etc.). The template identifies these types of documents. - As discussed above, the classifier also receives the harvested documents. The
classifier 106 extracts an activity type from each harvested document, much in the same way that theclassifier 106 extracts the activity type from the triggering item. In one embodiment, theclassifier 106 is a maximum entropy classifier. Theclassifier 106 is coupled to theset builder 108 and provides the selected template and the document activity types to the set builder. - The
set builder 108 is coupled to theFTQ processor 104 and to theclassifier 106 and receives the list of ranked documents and the selected template from theFTQ processor 104 and theclassifier 106, respectively. Theset builder 108 uses the. selected template and the activity types to re-rank and filter the list of ranked documents. In particular, theset builder 108 re-ranks the documents based on those whose activity type most closely matches the activity type of the selected template. Theset builder 108 outputs a final set of documents relevant to the triggering item. In one embodiment, theset builder 108 receives user feedback regarding the final set of documents. The user feedback may be used to refine the assembly of future sets of documents, as discussed in further detail below. - Those skilled in the art will appreciate that
FIG. 1 illustrates only one possible configuration of thesystem 100. In alternative configurations, for example, two or more of the individual system components could be replaced by a single component. Conversely, any single one of the system components could be replaced by multiple components. -
FIG. 2 is a schematic diagram illustrating a second embodiment of asystem 200 for assembling a set of documents related to a triggering item, according to the present invention. Like thesystem 100, thesystem 200 is implemented within a computing device that is operated by a user and that may or may not be connected to a network. - As above, the inputs to the
system 200 comprise a set of documents and a triggering item. In turn, the system outputs a subset of the documents that is relevant to the triggering item. As illustrated, thesystem 200 comprises four main components: aharvester 202, a full text query (FTQ)processor 204, apipeline 206 comprising a plurality of processors 210 1-210 n (hereinafter collectively referred to as “processors 210”), and aset builder 208. - The
harvester 202 indexes and processes documents including files, calendar events, to do lists, notes, emails, and email attachments. These documents may be retrieved locally from the user's computer and/or remotely from network storage. In the latter case, theharvester 202 may also retrieve documents from the World Wide Web (e.g., web pages). Theharvester 202 is coupled to theFTQ processor 204 and thepipeline 206, such that theharvester 202 may provide harvested documents to theFTQ processor 204 and thepipeline 206. (0026) TheFTQ processor 204, as discussed above, is coupled to theharvester 202. In addition to the harvested documents, the FTQ processor also receives triggering items from the computing device. As discussed above, a triggering item may comprise an event or a document. The triggering item may be entered directly by the user or may be entered automatically by an application running on the computing device. - The
FTQ processor 204 estimates the relevancy of the triggering item to the harvested documents. In one embodiment, the FTQ processor extracts features from the triggering item (such as keywords, people, dates, locations, acronyms, or the like) and then builds a query around these features. In one embodiment, the features are assigned weights in the query. The query is then run against a search (e.g., a LUCENE full text query) over the harvested documents. This produces a set of raw data that can be used to relate the harvested documents to the triggering item. In an alternative embodiment, the FTQ processor may also rank the harvested documents based upon the features (words) and document frequency. TheFTQ processor 204 is coupled to thepipeline 206 and provides the raw data (or the ranked list) to thepipeline 206. - The
pipeline 206 also receives the triggering item. Thepipeline 206 uses information about the triggering item and the raw data (or the ranked list of documents) produced by theFTQ processor 204 to filter and rank (or re-rank) the documents. As discussed above, thepipeline 206 comprises a plurality of pipelined processors 210 (i.e., the output of oneprocessor 210 passes as input to the next processor in the series). In one embodiment, each of theprocessors 210 contains its own persistent state (model, classifier, etc.). Theprocessors 210 are arranged such that thefirst processor 210 1 establishes an initial set of scores or rankings for the documents, and the subsequent processors 210 2-210 n adjust these scores. Each of theprocessors 210 may adjust the scores based on a different algorithm or model. For example, theprocessors 210 adjust the scores based on text frequency/inverse document frequency, proximity, or object recency, among other criteria. Thepipeline 206 is coupled to theset builder 208 and provides the list of filtered and re-ranked documents to the set builder. - The
set builder 208 is coupled to thepipeline 206 and receives the list of filtered and ranked documents from thepipeline 206. Theset builder 208 outputs a final set of documents relevant to the triggering item. - In one embodiment, the
system 200 additionally comprises a feedback bus 212 that receives user feedback regarding the final set of documents. The feedback bus is coupled to at least one of theprocessors 210 in thepipeline 206 and to theset builder 208. In this way, theset builder 208 or any coupledprocessor 210 may access the user feedback that is available via the feedback bus 212. The user feedback may be used to refine the assembly of future sets of documents. For example, the user feedback may be provided to any one or more of the processors 210 (e.g., automatically or responsive to a request from one or more of the processors 210). Where theprocessors 210 are capable of maintaining their own persistent states, the user feedback can be used to help train or adjust the algorithms or models used by theprocessors 210. - Those skilled in the art will appreciate that
FIG. 2 illustrates only one possible configuration of thesystem 200. In alternative configurations, for example, two or more of the individual system components could be replaced by a single component. Conversely, any single one of the system components could be replaced by multiple components. -
FIG. 3 is a flow diagram illustrating one embodiment of amethod 300 for assembling a set of documents related to a triggering item, according to the present invention. Themethod 300 may be implemented, for example, by thesystems FIGS. 1 and 2 . As such, reference is made in the discussion ofFIG. 3 to various elements ofFIGS. 1 and 2 . It will be appreciated, however, that themethod 300 is not limited to execution within a system configured exactly as illustrated inFIG. 1 orFIG. 2 and, may, in fact, execute within systems having alternative configurations. - The
method 300 is initialized atstep 302 and proceeds to step 304, where the harvester 102 (or 202) harvests documents including files (e.g., word processing files, spreadsheet files, presentation files, individual slides in presentation files, etc.), calendar events, to do lists, notes, emails, and email attachments. These documents may be retrieved locally from the user's computer and/or remotely from network storage. In the latter case, the harvested documents may also include documents retrieved from the World Wide Web (e.g., web pages). - In
step 306, theFTQ processor 104 and the classifier 106 (orFTQ processor 204 and pipeline 206) receive a triggering item (e.g., a document or event). In one embodiment, the system 100 (or 200) proactively detects the triggering item and assembles the documents (i.e., without receiving a request from the user to assemble the documents). For example, the system 100 (or 200) may detect the addition of a new item to the user's calendar and automatically assemble a set of relevant documents. In yet another embodiment, the triggering event comprises the user accessing an application that assembles the set of documents. - In
step 308, theclassifier 106 extracts an activity type from the triggering item and from each of the harvested documents. It should be noted that the activity type may be extracted from the harvested documents in advance of the receipt of the triggering item (i.e., the activity type is not necessarily extracted from the harvested documents simultaneously with the extraction of the activity type from the triggering item). For instance, in one embodiment, harvesting and classification of documents is performed substantially continuously (e.g., on a long-term basis), even when there are no triggering items on which to operate. In one embodiment, the activity type is determined by first parsing the text of the triggering item or the document into separate words (removing punctuation and stop words). An information or semantic extraction algorithm may be used to find word types. Theclassifier 106 then determines the activity type by classifying the triggering item or the document in accordance with the parsed words. In one embodiment, the classification is performed in accordance with a maximum entropy classifier. In another embodiment (e.g., where themethod 300 executes within thesystem 200 illustrated inFIG. 2 , one or more of theprocessors 210 may extract the activity type). - In
step 310, the FTQ processor 104 (or 204) builds a query in accordance with features extracted from the triggering item. The features may be, but are not necessarily, the same as the words that are parsed for classification purposes. In one embodiment, the extracted features include at least one of: a keyword, a person, a date, a location, or an acronym. Stop words are discarded. In one embodiment, the query is built by assigning weights to the extracted features. In one embodiment, the query is a Lucene full text query. - In
step 312, the FTQ processor 104 (or 204) runs the query against a search over the harvested documents. This produces a list of ranked documents based upon the extracted features (words) and document frequency. - In
step 314, theclassifier 106 selects a template in accordance with the activity type of the triggering item. The set oftemplates 110 includes a plurality of templates associated with different activity types. A given template identifies, for the associated activity type, the kinds of documents that a user would typically wish to gather. For example, a template for the activity type “business trip” might list the following types of documents: flight information, hotel reservation, car reservation, weather forecast, contact information, and the like. In one embodiment, thesystem 100 is pre-trained with knowledge of certain activity types and their associated templates; however, thesystem 100 can also be trained to learn new activity types and their associated templates. In one embodiment, theclassifier 106 selects the appropriate template by using the triggering item's activity type as a key into a template hash table. - In
step 316, theset builder 108 re-ranks and filters the list of ranked documents produced by theFTQ processor 104, in accordance with the template selected by theclassifier 106 and the activity types of the documents in the list of ranked documents. In one embodiment, theset builder 108 does this by comparing the activity type of each document to the activity type of the selected template. The ranks of the documents whose activity types match the activity type of the template are increased, while the ranks of the documents whose activity types do not match the activity type of the template are reduced (or the documents are removed from the list). In the case of thesystem 200, the re-ranking and filtering is performed by thepipeline 206, in accordance with the algorithms or models implemented in each of theprocessors 210. In this instance, it is possible that no templates may be used. - In one embodiment, ranking the documents includes recency ranking, word popularity, and learning. An initial ranking may be based on a combination of recency ranking and word popularity. Recency ranking assigns a higher weight to documents having later modification dates, access dates, and/or number of accesses. Word popularity uses a global dictionary containing high frequency words and weights. The number of high frequency words is user-configurable and may be set to a finite number of words (e.g., fifty thousand). When a new document is harvested, keywords that do not exist in the dictionary are added to the dictionary.
- In a further embodiment, stemming is applied to obtain root words, and all words are converted to the lowercase letters. Weights in the dictionary may be updated continuously as the system 100 (or 200) operates. Keywords appearing in frequently accesses documents may be assigned higher weights, while keywords appearing in less frequently accessed documents may be assigned lower weights. In this way, the system 100 (or 200) has access to an up-to-date and accurate model of the user's universe and behavior.
- The re-ranking and filtering performed in
step 316 produces a final set of documents. The final set of documents comprises at least one document that the system 100 (or 200) believes is relevant to the triggering item. In one embodiment, the final set of documents may also include an alert to notify the user of potentially missing information (e.g., no hotel reservation was located for a business trip). Alternatively, the system 100 (or 200) could interface with a task learning application that determines the appropriate procedure to execute to retrieve the missing document. The final set of documents is output by the set builder 108 (or 208) instep 318. - In
step 320, theset builder 108 optionally receives user feedback regarding the final set of documents. The user feedback may be implicit or explicit. Explicit feedback might include, for example, a user command to keep a given document in the final set of documents or to remove a given document from the final set of documents. Implicit feedback might include, for example, the fact the a user did or did not open a given document in the final set of documents, that the user opened a folder containing a given document in the final set of documents, or asked for items that are similar to a given document in the final set of documents. - In
step 322, the set builder 108 (or 208) stores the user feedback. Storage of the user feedback involves adjusting the way in which theset builder 108 builds sets of documents. In one embodiment, the set builder 108 (or 208) updates the global dictionary (discussed above) at the word level. For example, the weights of the most significant words in the document corresponding to the user feedback may be increased if the user feedback is positive, or decreased if the user feedback is negative. Additionally, keywords from the query built instep 310 may be added to the document as metadata. If the keywords also exist in the global dictionary, their weights may also be adjusted as described above. The global dictionary may then be used to re-rank sets of documents assembled in response to future queries. The metadata stored with the documents may also be used to increase the rank of the documents (e.g., when the metadata matches future query keywords, the documents' ranks are increased). - In another embodiment, the set builder 108 (or 208) makes updates at the document level. For example, each document in the final set of documents may be associated with a signature of the triggering item, where the signature also indicates whether the user feedback with respect to the document was positive or negative. If the document then shows up in a future list of ranked documents, the set builder 108 (or 208) will compare the signature to the signature of the current triggering item. If the signatures match and the user feedback was positive, the document's ranking is increased; if the signatures match and the user feedback was negative, the document's ranking is decreased. In one embodiment, explicit feedback has a more significant effect on ranking than implicit feedback.
- In the case of the
system 200, theset builder 208 may provide the user feedback to one or more of theprocessors 210. As discussed above, any of theprocessors 210 may adjust the algorithms or models implemented therein in accordance with the user feedback. - The
method 300 terminates instep 324. - As discussed above, the
system 100 may be pre-trained with knowledge of certain activity types and their associated templates. However, thesystem 100 can also be trained by the user to learn other activity types and their associated templates. To do this, the user identifies a new document and its associated activity type. This information is fed to theclassifier 106, which subsequently parses the new document for words and word types (e.g., dates, times, names, etc.). An information or semantic extraction algorithm may be used. The activity type, words, and word types are then stored by theclassifier 106 as a new class. - In a similar manner, the
system 100 can also be trained to recognize new documents of a known activity type. A new document is parsed into words and word types as described above. These words and word types are used, along with the known activity type, as training data for theclassifier 106. - As discussed above, the
systems method 300 may be deployed in substantially any computing device or network. Although the examples discussed above relate mostly to desk top applications, other applications are envisioned. For example, thesystem system system system - In further embodiments still, the
system system - Sorting and/or ranking of this information may be based on the time of day and/or the user's interest level (e.g., as expressed in a user profile). For instance, if the current time of day falls within the user's normal working hours, work-related information (e.g., “John just scheduled a meeting for 2:00 P.M. tomorrow”) might be ranked more highly than information that is not work related (e.g., “Tickets go on sale for the Giants game in ten minutes”). The type of information provided, as well as the manner and time in which the information is provided, may by learned dynamically based on user feedback or based on a user profile.
- In one embodiment, the set of personalized documents is assembled on a periodic basis (e.g., every thirty minutes). In another embodiment, the set of personalized documents is assembled as new relevant information is detected. Time sensitive documents may be removed from the set as they become outdated, so that the storage required by the
system system - The set of personalized documents may also include hyperlinks or other executable code that allows the user to share one or more documents in an email or on a social networking site (e.g., email a family member to let them know that a flight has been delayed), to add an event to the user's calendar (e.g., add an entry for a child's upcoming soccer game), to initiate a commercial transaction (e.g., purchase tickets to an upcoming concert), or to simply obtain more information about a given document (e.g., read the full news article about a particular event).
- Thus, the set of personalized documents might resemble the following:
-
- It is 11:45 A.M., cloudy, and 76 degrees Fahrenheit
- Your son scheduled a soccer game on Saturday at 2:00 P.M.
- Your flight to Denver has been delayed by 50 minutes
- Brad Paisley is playing a concert nearby next week (Buy tickets?)
- The Giants won (Email your friends?)
- You have two new email messages from your boss
- In another embodiment, this information may be grouped and/or ranked. In this case, the above set of personalized documents might resemble the following:
- It is 11:45 A.M., cloudy, and 76 degrees Fahrenheit
- Your flight to Denver has been delayed by 50 minutes
- You have two new email messages from your boss
- Your son scheduled a soccer game on Saturday at 2:00 P.M.
- Brad Paisley is playing a concert nearby next week (Buy tickets?)
- The Giants won (Email your friends?)
-
FIG. 4 is a high level block diagram of the present invention implemented using a generalpurpose computing device 400. It should be understood that embodiments of the invention can be implemented as a physical device or subsystem that is coupled to a processor through a communication channel. Therefore, in one embodiment, a generalpurpose computing device 400 comprises aprocessor 402, amemory 404, adocument assembly module 405, and various input/output (I/O)devices 406 such as a display, a keyboard, a mouse, a modem, a microphone, speakers, a touch screen, and the like. In one embodiment, at least one I/O device is a storage device (e.g., a disk drive, an optical disk drive, a flash drive). - Alternatively, embodiments of the present invention (e.g., document assembly module 405) can be represented by one or more software applications (or even a combination of software and hardware, e.g., using Application Specific Integrated Circuits (ASIC)), where the software is loaded from a storage medium (e.g., I/O devices 406) and operated by the
processor 402 in thememory 404 of the generalpurpose computing device 400. Thus, in one embodiment, thedocument assembly module 405 for assembling a set of documents related to a triggering item described herein with reference to the preceding Figures can be stored on a non-transitory computer readable medium (e.g., RAM, magnetic or optical drive or diskette, and the like). - It should be noted that although not explicitly specified, one or more steps of the methods described herein may include a storing, displaying and/or outputting step as required for a particular application. In other words, any data, records, fields, and/or intermediate results discussed in the methods can be stored, displayed, and/or outputted to another device as required for a particular application. Furthermore, steps or blocks in the accompanying Figures that recite a determining operation or involve a decision, do not necessarily require that both branches of the determining operation be practiced. In other words, one of the branches of the determining operation can be deemed as an optional step.
- Although various embodiments which incorporate the teachings of the present invention have been shown and described in detail herein, those skilled in the art can readily devise many other varied embodiments that still incorporate these teachings.
Claims (30)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US13/182,245 US9904681B2 (en) | 2009-01-12 | 2011-07-13 | Method and apparatus for assembling a set of documents related to a triggering item |
Applications Claiming Priority (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14411709P | 2009-01-12 | 2009-01-12 | |
US21925909P | 2009-06-22 | 2009-06-22 | |
US12/632,491 US20100180200A1 (en) | 2009-01-12 | 2009-12-07 | Electronic assistant |
US41572210P | 2010-11-19 | 2010-11-19 | |
US13/182,245 US9904681B2 (en) | 2009-01-12 | 2011-07-13 | Method and apparatus for assembling a set of documents related to a triggering item |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US12/632,491 Continuation-In-Part US20100180200A1 (en) | 2009-01-12 | 2009-12-07 | Electronic assistant |
Publications (3)
Publication Number | Publication Date |
---|---|
US20120131020A1 US20120131020A1 (en) | 2012-05-24 |
US20160188578A9 true US20160188578A9 (en) | 2016-06-30 |
US9904681B2 US9904681B2 (en) | 2018-02-27 |
Family
ID=46065342
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US13/182,245 Active 2030-02-13 US9904681B2 (en) | 2009-01-12 | 2011-07-13 | Method and apparatus for assembling a set of documents related to a triggering item |
Country Status (1)
Country | Link |
---|---|
US (1) | US9904681B2 (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9519461B2 (en) | 2013-06-20 | 2016-12-13 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on third-party developers |
US9594542B2 (en) | 2013-06-20 | 2017-03-14 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on training by third-party developers |
US9633317B2 (en) | 2013-06-20 | 2017-04-25 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on a natural language intent interpreter |
US10474961B2 (en) | 2013-06-20 | 2019-11-12 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on prompting for additional user input |
Families Citing this family (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9443007B2 (en) | 2011-11-02 | 2016-09-13 | Salesforce.Com, Inc. | Tools and techniques for extracting knowledge from unstructured data retrieved from personal data sources |
US9471666B2 (en) | 2011-11-02 | 2016-10-18 | Salesforce.Com, Inc. | System and method for supporting natural language queries and requests against a user's personal data cloud |
US8832057B2 (en) * | 2011-12-02 | 2014-09-09 | Yahoo! Inc. | Results returned for list-seeking queries |
US9046917B2 (en) | 2012-05-17 | 2015-06-02 | Sri International | Device, method and system for monitoring, predicting, and accelerating interactions with a computing device |
US10163058B2 (en) | 2012-08-14 | 2018-12-25 | Sri International | Method, system and device for inferring a mobile user's current context and proactively providing assistance |
US9015099B2 (en) | 2012-08-14 | 2015-04-21 | Sri International | Method, system and device for inferring a mobile user's current context and proactively providing assistance |
US9374422B2 (en) | 2012-12-18 | 2016-06-21 | Arash Esmailzadeh | Secure distributed data storage |
US11074618B2 (en) | 2013-06-13 | 2021-07-27 | Blackberry Limited | Method and apparatus pertaining to history-based content-sharing recommendations |
US10049342B2 (en) | 2013-06-13 | 2018-08-14 | Blackberry Limited | Method and apparatus pertaining to sharing content with scheduled-event participants |
US20150026002A1 (en) * | 2013-07-16 | 2015-01-22 | International Business Machines Corporation | Smart home environment for shopping |
US10367649B2 (en) | 2013-11-13 | 2019-07-30 | Salesforce.Com, Inc. | Smart scheduling and reporting for teams |
US9893905B2 (en) | 2013-11-13 | 2018-02-13 | Salesforce.Com, Inc. | Collaborative platform for teams with messaging and learning across groups |
WO2015079354A1 (en) * | 2013-11-26 | 2015-06-04 | Koninklijke Philips N.V. | System and method of determining missing interval change information in radiology reports |
US10242069B2 (en) * | 2015-03-13 | 2019-03-26 | Microsoft Technology Licensing, Llc | Enhanced template curating |
US9762520B2 (en) | 2015-03-31 | 2017-09-12 | Salesforce.Com, Inc. | Automatic generation of dynamically assigned conditional follow-up tasks |
CN104852842A (en) * | 2015-04-29 | 2015-08-19 | 小米科技有限责任公司 | Information broadcasting method and information broadcasting device |
US11227261B2 (en) | 2015-05-27 | 2022-01-18 | Salesforce.Com, Inc. | Transactional electronic meeting scheduling utilizing dynamic availability rendering |
US10592541B2 (en) * | 2015-05-29 | 2020-03-17 | Intel Corporation | Technologies for dynamic automated content discovery |
US11477302B2 (en) | 2016-07-06 | 2022-10-18 | Palo Alto Research Center Incorporated | Computer-implemented system and method for distributed activity detection |
US10431188B1 (en) * | 2016-08-19 | 2019-10-01 | Amazon Technologies, Inc. | Organization of personalized content |
US11200282B1 (en) | 2018-03-22 | 2021-12-14 | Atlassian Pty Ltd. | Integrated views of multiple different computer program applications with action options |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6480830B1 (en) * | 1998-01-29 | 2002-11-12 | International Business Machines Corporation | Active calendar system |
US20030115189A1 (en) * | 2001-12-19 | 2003-06-19 | Narayan Srinivasa | Method and apparatus for electronically extracting application specific multidimensional information from documents selected from a set of documents electronically extracted from a library of electronically searchable documents |
US6640230B1 (en) * | 2000-09-27 | 2003-10-28 | International Business Machines Corporation | Calendar-driven application technique for preparing responses to incoming events |
US20080301190A1 (en) * | 2007-06-01 | 2008-12-04 | Kendall Gregory Lockhart | Integrated system and method for implementing messaging, planning, and search functions in a mobile device |
US20090100347A1 (en) * | 2005-12-21 | 2009-04-16 | Schemers Roland J | Enhanced calendaring functionality in a collaboration suite |
US20100070877A1 (en) * | 2008-09-17 | 2010-03-18 | Microsoft Corporation | Seamless conversion of ordinary email data into calendar data |
US20100094839A1 (en) * | 2008-10-14 | 2010-04-15 | Brach Ricci S | Coordinated notification |
US7797331B2 (en) * | 2002-12-20 | 2010-09-14 | Nokia Corporation | Method and device for organizing user provided information with meta-information |
US7831601B2 (en) * | 2004-08-04 | 2010-11-09 | International Business Machines Corporation | Method for automatically searching for documents related to calendar and email entries |
US7970637B2 (en) * | 2006-06-27 | 2011-06-28 | Microsoft Corporation | Activity-centric granular application functionality |
US8266148B2 (en) * | 2008-10-07 | 2012-09-11 | Aumni Data, Inc. | Method and system for business intelligence analytics on unstructured data |
Family Cites Families (30)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE69523543T2 (en) | 1994-12-13 | 2002-04-04 | Microsoft Corp | Taskbar with start menu |
JP3664874B2 (en) | 1998-03-28 | 2005-06-29 | 松下電器産業株式会社 | Document search device |
US7076546B1 (en) | 1999-02-10 | 2006-07-11 | International Business Machines Corporation | Browser for use in accessing hypertext documents in a multi-user computer environment |
US7343365B2 (en) * | 2002-02-20 | 2008-03-11 | Microsoft Corporation | Computer system architecture for automatic context associations |
US7587411B2 (en) | 2003-03-27 | 2009-09-08 | Microsoft Corporation | System and method for filtering and organizing items based on common elements |
US7912820B2 (en) * | 2003-06-06 | 2011-03-22 | Microsoft Corporation | Automatic task generator method and system |
US7836010B2 (en) * | 2003-07-30 | 2010-11-16 | Northwestern University | Method and system for assessing relevant properties of work contexts for use by information services |
US7457872B2 (en) * | 2003-10-15 | 2008-11-25 | Microsoft Corporation | On-line service/application monitoring and reporting system |
US7747621B2 (en) | 2003-12-19 | 2010-06-29 | Canon Kabushiki Kaisha | Generating a relation diagram of data files |
US8099407B2 (en) * | 2004-03-31 | 2012-01-17 | Google Inc. | Methods and systems for processing media files |
US7634461B2 (en) * | 2004-08-04 | 2009-12-15 | International Business Machines Corporation | System and method for enhancing keyword relevance by user's interest on the search result documents |
US7493303B2 (en) * | 2004-08-04 | 2009-02-17 | International Business Machines Corporation | Method for remotely searching a local user index |
US7496567B1 (en) | 2004-10-01 | 2009-02-24 | Terril John Steichen | System and method for document categorization |
US7523137B2 (en) * | 2005-04-08 | 2009-04-21 | Accenture Global Services Gmbh | Model-driven event detection, implication, and reporting system |
WO2007052285A2 (en) * | 2005-07-22 | 2007-05-10 | Yogesh Chunilal Rathod | Universal knowledge management and desktop search system |
US7689933B1 (en) | 2005-11-14 | 2010-03-30 | Adobe Systems Inc. | Methods and apparatus to preview content |
US9177124B2 (en) * | 2006-03-01 | 2015-11-03 | Oracle International Corporation | Flexible authentication framework |
US7941419B2 (en) * | 2006-03-01 | 2011-05-10 | Oracle International Corporation | Suggested content with attribute parameterization |
US7756898B2 (en) * | 2006-03-31 | 2010-07-13 | Isilon Systems, Inc. | Systems and methods for notifying listeners of events |
US8364514B2 (en) | 2006-06-27 | 2013-01-29 | Microsoft Corporation | Monitoring group activities |
US7739622B2 (en) | 2006-10-27 | 2010-06-15 | Microsoft Corporation | Dynamic thumbnails for document navigation |
US8056007B2 (en) * | 2006-11-15 | 2011-11-08 | Yahoo! Inc. | System and method for recognizing and storing information and associated context |
US20080126983A1 (en) | 2006-11-29 | 2008-05-29 | Keohane Susann M | Content-based ordering of a list of selectable entries for an auto-complete box |
US8468244B2 (en) * | 2007-01-05 | 2013-06-18 | Digital Doors, Inc. | Digital information infrastructure and method for security designated data and with granular data stores |
US20090199133A1 (en) | 2008-02-05 | 2009-08-06 | Microsoft Corporation | Generating a destination list utilizing usage data |
US8442940B1 (en) * | 2008-11-18 | 2013-05-14 | Semantic Research, Inc. | Systems and methods for pairing of a semantic network and a natural language processing information extraction system |
US8862691B2 (en) | 2008-12-22 | 2014-10-14 | Microsoft Corporation | Media aggregation and presentation |
US20100180200A1 (en) | 2009-01-12 | 2010-07-15 | Thierry Donneau-Golencer | Electronic assistant |
US8407078B1 (en) * | 2009-01-20 | 2013-03-26 | Perot Systems Corporation | Method of and system for managing projects, programs and portfolios throughout the project lifecycle |
US8341534B2 (en) * | 2010-03-05 | 2012-12-25 | Palo Alto Research Center Incorporated | System and method for flexibly taking actions in response to detected activities |
-
2011
- 2011-07-13 US US13/182,245 patent/US9904681B2/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6480830B1 (en) * | 1998-01-29 | 2002-11-12 | International Business Machines Corporation | Active calendar system |
US6640230B1 (en) * | 2000-09-27 | 2003-10-28 | International Business Machines Corporation | Calendar-driven application technique for preparing responses to incoming events |
US20030115189A1 (en) * | 2001-12-19 | 2003-06-19 | Narayan Srinivasa | Method and apparatus for electronically extracting application specific multidimensional information from documents selected from a set of documents electronically extracted from a library of electronically searchable documents |
US7797331B2 (en) * | 2002-12-20 | 2010-09-14 | Nokia Corporation | Method and device for organizing user provided information with meta-information |
US7831601B2 (en) * | 2004-08-04 | 2010-11-09 | International Business Machines Corporation | Method for automatically searching for documents related to calendar and email entries |
US20090100347A1 (en) * | 2005-12-21 | 2009-04-16 | Schemers Roland J | Enhanced calendaring functionality in a collaboration suite |
US7970637B2 (en) * | 2006-06-27 | 2011-06-28 | Microsoft Corporation | Activity-centric granular application functionality |
US20080301190A1 (en) * | 2007-06-01 | 2008-12-04 | Kendall Gregory Lockhart | Integrated system and method for implementing messaging, planning, and search functions in a mobile device |
US20100070877A1 (en) * | 2008-09-17 | 2010-03-18 | Microsoft Corporation | Seamless conversion of ordinary email data into calendar data |
US8266148B2 (en) * | 2008-10-07 | 2012-09-11 | Aumni Data, Inc. | Method and system for business intelligence analytics on unstructured data |
US20100094839A1 (en) * | 2008-10-14 | 2010-04-15 | Brach Ricci S | Coordinated notification |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9519461B2 (en) | 2013-06-20 | 2016-12-13 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on third-party developers |
US9594542B2 (en) | 2013-06-20 | 2017-03-14 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on training by third-party developers |
US9633317B2 (en) | 2013-06-20 | 2017-04-25 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on a natural language intent interpreter |
US10083009B2 (en) | 2013-06-20 | 2018-09-25 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system planning |
US10474961B2 (en) | 2013-06-20 | 2019-11-12 | Viv Labs, Inc. | Dynamically evolving cognitive architecture system based on prompting for additional user input |
Also Published As
Publication number | Publication date |
---|---|
US9904681B2 (en) | 2018-02-27 |
US20120131020A1 (en) | 2012-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9904681B2 (en) | Method and apparatus for assembling a set of documents related to a triggering item | |
EP3819792A2 (en) | Method, apparatus, device, and storage medium for intention recommendation | |
US20100180200A1 (en) | Electronic assistant | |
US20200110785A1 (en) | Personalized search filter and notification system | |
CN109690529B (en) | Compiling documents into a timeline by event | |
US9830386B2 (en) | Determining trending topics in social media | |
US8868558B2 (en) | Quote-based search | |
US8311997B1 (en) | Generating targeted paid search campaigns | |
US9443226B2 (en) | Electronic assistant for making predictions based on user messages | |
US7809664B2 (en) | Automated learning from a question and answering network of humans | |
US8805823B2 (en) | Content processing systems and methods | |
US20170161375A1 (en) | Clustering documents based on textual content | |
US8793259B2 (en) | Information retrieval device, information retrieval method, and program | |
US10110544B2 (en) | Method and system for classifying a question | |
US20130297582A1 (en) | Peer sharing of personalized views of detected information based on relevancy to a particular user's personal interests | |
US9633119B2 (en) | Content ranking based on user features in content | |
US20110295612A1 (en) | Method and apparatus for user modelization | |
JP2006107473A (en) | System, method, and interface for providing personalized search and information access | |
US20110219299A1 (en) | Method and system of providing completion suggestion to a partial linguistic element | |
US20160085869A1 (en) | Social media content analysis and output | |
EP2524348A2 (en) | User communication analysis systems and methods | |
WO2011062598A1 (en) | System and method for automated filtering of reviews for marketability | |
JP2016126567A (en) | Content recommendation device and program | |
US11640420B2 (en) | System and method for automatic summarization of content with event based analysis | |
US20170098168A1 (en) | Method and system for providing synthetic answers to a personal question |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: SRI INTERNATIONAL, CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:NITZ, KENNETH;DUNKELY, DAVID;DONNEAU-GOLENCER, THIERRY;AND OTHERS;SIGNING DATES FROM 20110714 TO 20110830;REEL/FRAME:027162/0399 |
|
AS | Assignment |
Owner name: AFRL/RIJ, NEW YORK Free format text: CONFIRMATORY LICENSE;ASSIGNOR:SRI INTERNATIONAL;REEL/FRAME:027364/0496 Effective date: 20111201 |
|
FEPP | Fee payment procedure |
Free format text: ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.) |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 4 |