TWI493364B - Management methods for subjective comments of articles, and related devices and computer program products - Google Patents
Management methods for subjective comments of articles, and related devices and computer program products Download PDFInfo
- Publication number
- TWI493364B TWI493364B TW102118215A TW102118215A TWI493364B TW I493364 B TWI493364 B TW I493364B TW 102118215 A TW102118215 A TW 102118215A TW 102118215 A TW102118215 A TW 102118215A TW I493364 B TWI493364 B TW I493364B
- Authority
- TW
- Taiwan
- Prior art keywords
- article
- attribute
- word
- sentence
- opinion
- Prior art date
Links
- 238000007726 management method Methods 0.000 title claims description 41
- 238000004590 computer program Methods 0.000 title claims description 6
- 238000012549 training Methods 0.000 claims description 55
- 238000011156 evaluation Methods 0.000 claims description 33
- 238000012545 processing Methods 0.000 claims description 18
- 238000000034 method Methods 0.000 claims description 13
- 238000012706 support-vector machine Methods 0.000 description 17
- 230000006870 function Effects 0.000 description 4
- 230000008451 emotion Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013518 transcription Methods 0.000 description 1
- 230000035897 transcription Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Description
本發明係有關於一種文章之主觀意見管理方法及其相關裝置，且特別有關於一種可以對於網路文章之主觀意見進行識別之方法及其相關裝置。The present invention relates to a method for subjective opinion management of an article and related devices, and particularly relates to a method for identifying subjective opinions of a web article and related devices.
近年來，隨著各式各樣具有網路連接能力之電子裝置，如電腦、筆記型電腦、平板電腦、及智慧型手機的問市，使用者可以隨時隨地的利用電子裝置來連接網路，以瀏覽網際網路，且透過網路進行相關應用與服務。由於網路及這些裝置及其功能所帶來的便利，也使得這些裝置成為現代人的必備品之一，並隨時隨地的在生活中頻繁使用這些功能。In recent years, with the various electronic devices with network connectivity, such as computers, notebook computers, tablet computers, and smart phones, users can use the electronic devices to connect to the Internet anytime, anywhere. To browse the Internet and conduct related applications and services over the Internet. Due to the convenience of the network and these devices and their functions, these devices have become one of the must-have items for modern people, and they are frequently used in life anytime and anywhere.
同時，隨著網際網路的發展，網路經營者積極地開發各式各樣的網路應用給使用者使用。舉例來說，使用者可以透過電子佈告欄系統(BBS)來瀏覽文章與發表評論。使用者也可以透過特定網站或是自行架設網頁來發表文章。目前，網際網路上正盛行的是部落格(Blog)的應用。每一個使用者可以建立自己的部落格，在自己在部落格中發表文章。At the same time, with the development of the Internet, network operators are actively developing a variety of network applications for users to use. For example, users can browse articles and post comments through the bulletin board system (BBS). Users can also post articles through specific websites or by setting up their own web pages. At present, the application of blogs is prevalent on the Internet. Each user can create their own blog and post an article in the blog.
由於部落格及其他類似網路平台的興起，在任一網路文章中必定會有一至多個在描述或談論的物件，此物件可以係人物、事件、時間、地點、物品等。文章作者可以對於物件本身表示意見，也可以對於物件寫下評論描述文字，從而產生使用者對於物件之主觀性意見。相對於相關廠商，其也期望透過其上發佈的網路文章來得知使用者對於其商品或服務的消費意見，以進行相關改進或廣告行為。Due to the rise of blogs and other similar web platforms, there must be one or more objects in the description or discussion in any web article, which can be characters, events, time, places, items, and so on. The author of the article can express opinions on the object itself, and can also write a comment description text on the object, thereby generating the user's subjective opinion on the object. Relative to the relevant vendors, they also expect to know the user's opinions on the consumption of their goods or services through the online articles published on them to make relevant improvements or advertising behaviors.
習知地，網路文章的分析技術僅限於單純利用事先定義的特定詞庫，如屬性詞庫、意見詞庫、程度詞庫、否定詞庫等來比對網路文 章，從而取得相關物件之屬性及其對應之意見詞。然而，習知文章分析的結果往往不如預期。舉例來說，習知技術由網路文章找出之主觀意見往往不夠正確。因此，目前業界亟需一種可以強化主觀意見之識別，與/或加強網路文章與詞庫比對效率之技術。Conventionally, the analysis technique of online articles is limited to the use of specific lexicons defined in advance, such as attribute lexicon, opinion lexicon, degree vocabulary, negative lexicon, etc. Chapter, in order to obtain the attributes of related objects and their corresponding opinions. However, the results of conventional article analysis are often not as expected. For example, the subjective opinions of conventional techniques found by online articles are often not correct enough. Therefore, there is a need in the industry for a technique that enhances the identification of subjective opinions and/or enhances the efficiency of online articles and thesaurus.
有鑑於此，本發明提供文章之主觀意見管理方法及其相關裝置，其中，電子裝置可以對於網路文章進行主觀意見之識別作業。In view of this, the present invention provides a subjective opinion management method and related apparatus of the article, wherein the electronic device can perform subjective opinion recognition work on the web article.
本發明實施例之一種文章之主觀意見管理方法。首先，取得一文章語句。依據包括複數候選屬性詞之一屬性詞庫搜尋文章語句中之一屬性詞，並將屬性詞輸入一距離模型，從而得到相應屬性詞之一標準距離值。接著，由文章語句取得一意見片語，且計算屬性詞與意見片語於文章語句中之一實際距離值。之後，依據標準距離值與實際距離值判斷文章語句是否為一非主觀意見語句。A method for subjective opinion management of an article in an embodiment of the present invention. First, get an article statement. Searching for one of the attribute words in the article sentence according to one of the attribute candidate words of the plurality of candidate attribute words, and inputting the attribute word into a distance model, thereby obtaining a standard distance value of the corresponding attribute word. Next, an opinion phrase is obtained from the article sentence, and an actual distance value of the attribute word and the opinion piece in the article sentence is calculated. Then, based on the standard distance value and the actual distance value, it is determined whether the article sentence is a non-subjective opinion statement.
本發明實施例之一種文章之主觀意見管理裝置至少包括一儲存單元、與一處理單元。儲存單元具有一文章語句、包括複數候選屬性詞之一屬性詞庫、與一距離模型。處理單元依據屬性詞庫搜尋文章語句中之一屬性詞，並將屬性詞輸入距離模型，從而得到相應屬性詞之一標準距離值。處理單元由文章語句取得一意見片語，且計算屬性詞與意見片語於文章語句中之一實際距離值。處理單元依據標準距離值與實際距離值判斷文章語句是否為一非主觀意見語句。The subjective opinion management apparatus of an article of the embodiment of the present invention includes at least a storage unit and a processing unit. The storage unit has an article sentence, an attribute vocabulary including a plurality of candidate attribute words, and a distance model. The processing unit searches for one attribute word in the article sentence according to the attribute vocabulary, and inputs the attribute word into the distance model, thereby obtaining a standard distance value of one of the corresponding attribute words. The processing unit obtains a comment phrase from the article sentence, and calculates an actual distance value between the attribute word and the opinion piece in the article sentence. The processing unit determines whether the article statement is a non-subjective opinion statement according to the standard distance value and the actual distance value.
在一些實施例中，可以依據屬性詞於文章語句中之位置及標準距離值決定由文章語句搜尋意見片語之一搜尋起始位置。In some embodiments, the search start position of one of the article sentences may be determined by the position of the attribute word in the article sentence and the standard distance value.
在一些實施例中，意見片語可以係依據一評價詞庫、一程度詞庫、與一否定詞庫進行搜尋文章語句所得到。In some embodiments, the commentary may be obtained by searching for an article sentence based on an evaluation vocabulary, a degree vocabulary, and a negative vocabulary.
在一些實施例中，當實際距離值大於或小於標準距離值時，判定文章語句為一非主觀意見語句。在一些實施例中，當實際距離值大於標準距離值加一容許誤差值，或小於標準距離值減一容許誤差值時，判定文章語句為一非主觀意見語句。In some embodiments, when the actual distance value is greater than or less than the standard distance value, the article statement is determined to be a non-subjective opinion statement. In some embodiments, when the actual distance value is greater than the standard distance value plus an allowable error value, or less than the standard distance value minus one allowable error value, the article sentence is determined to be a non-subjective opinion statement.
在一些實施例中，距離模型可以係依據複數訓練語句產 生，其中每一訓練語句至少包括一訓練屬性詞與一訓練意見片語，且訓練屬性詞與訓練意見片語間之一距離被作為一特徵值以輸入距離模型。In some embodiments, the distance model can be based on a plurality of training statements. And each training sentence includes at least one training attribute word and one training opinion piece, and a distance between the training attribute word and the training opinion piece is used as a feature value to input the distance model.
本發明實施例之一種文章之主觀意見管理方法。首先，取得一文章語句。依據包括複數候選屬性詞之一屬性詞庫搜尋文章語句中之一屬性詞，並將屬性詞輸入一距離模型，從而得到相應屬性詞之一標準距離值。接著，依據屬性詞於文章語句中之位置及標準距離值決定由文章語句搜尋一意見片語之一搜尋起始位置。A method for subjective opinion management of an article in an embodiment of the present invention. First, get an article statement. Searching for one of the attribute words in the article sentence according to one of the attribute candidate words of the plurality of candidate attribute words, and inputting the attribute word into a distance model, thereby obtaining a standard distance value of the corresponding attribute word. Then, according to the position of the attribute word in the article sentence and the standard distance value, it is determined that the article sentence searches for a search starting position of one comment phrase.
本發明實施例之一種文章之主觀意見管理裝置至少包括一儲存單元、與一處理單元。儲存單元具有一文章語句、包括複數候選屬性詞之一屬性詞庫、與一距離模型。處理單元依據屬性詞庫搜尋文章語句中之一屬性詞，並將屬性詞輸入距離模型，從而得到相應屬性詞之一標準距離值。處理單元依據屬性詞於文章語句中之位置及標準距離值決定由文章語句搜尋一意見片語之一搜尋起始位置。The subjective opinion management apparatus of an article of the embodiment of the present invention includes at least a storage unit and a processing unit. The storage unit has an article sentence, an attribute vocabulary including a plurality of candidate attribute words, and a distance model. The processing unit searches for one attribute word in the article sentence according to the attribute vocabulary, and inputs the attribute word into the distance model, thereby obtaining a standard distance value of one of the corresponding attribute words. The processing unit determines, according to the position of the attribute word in the article sentence and the standard distance value, the search starting position of one of the comment words by the article sentence.
在一些實施例中，距離模型可以係依據複數訓練語句產生，其中每一訓練語句至少包括一訓練屬性詞與一訓練意見片語，且訓練屬性詞與訓練意見片語間之一距離被作為一特徵值以輸入距離模型。In some embodiments, the distance model may be generated according to a plurality of training sentences, wherein each training sentence includes at least one training attribute word and one training opinion phrase, and one distance between the training attribute word and the training opinion phrase is taken as one The eigenvalue is entered into the distance model.
本發明之文章之主觀意見管理方法及其相關裝置可以對於網路文章進行主觀意見之識別作業，從而增加由文章中擷取主觀意見之正確性，且/或增加由文章中搜尋意見片語之效率。The subjective opinion management method and related device of the article of the present invention can perform subjective opinion recognition work on the online article, thereby increasing the correctness of the subjective opinion extracted from the article, and/or increasing the search for the commentary in the article. effectiveness.
本發明上述方法可以透過程式碼方式存在。當程式碼被機器載入且執行時，機器變成用以實行本發明之裝置。The above method of the present invention can exist in a coded manner. When the code is loaded and executed by the machine, the machine becomes the means for practicing the invention.
為使本發明之上述目的、特徵和優點能更明顯易懂，下文特舉實施例，並配合所附圖示，詳細說明如下。The above described objects, features, and advantages of the invention will be apparent from the description and appended claims appended claims
100‧‧‧文章之主觀意見管理裝置100‧‧‧ subjective opinion management device
110‧‧‧儲存單元110‧‧‧ storage unit
111‧‧‧文章語句111‧‧‧ article statement
112‧‧‧屬性詞庫112‧‧‧Attribute vocabulary
113‧‧‧距離模型113‧‧‧ distance model
120‧‧‧處理單元120‧‧‧Processing unit
S210、S220、S230、S240‧‧‧步驟S210, S220, S230, S240‧‧‧ steps
S310、S320、…、S360‧‧‧步驟S310, S320, ..., S360‧‧ steps
S410、S420‧‧‧步驟S410, S420‧‧‧ steps
S510、S520、…、S550‧‧‧步驟S510, S520, ..., S550‧‧ steps
第1圖為一示意圖係顯示依據本發明實施例之文章之主觀意見管理裝置。BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic diagram showing a subjective opinion management apparatus for an article in accordance with an embodiment of the present invention.
第2圖為一流程圖係顯示依據本發明實施例之距離模型訓練方法。Fig. 2 is a flow chart showing a distance model training method according to an embodiment of the present invention.
第3圖為一流程圖係顯示依據本發明實施例之文章之主觀意見管理方 法。Figure 3 is a flow chart showing the subjective opinion management party of the article according to an embodiment of the present invention. law.
第4圖為一流程圖係顯示依據本發明實施例之意見片語搜尋方法。Fig. 4 is a flow chart showing a method for searching for opinions in accordance with an embodiment of the present invention.
第5圖為一流程圖係顯示依據本發明另一實施例之文章之主觀意見管理方法。Figure 5 is a flow chart showing a subjective opinion management method for an article according to another embodiment of the present invention.
第1圖顯示依據本發明實施例之文章之主觀意見管理裝置。如第1圖所示，依據本發明實施例之文章之主觀意見管理裝置100可以適用於一電子裝置，如電腦。文章之主觀意見管理裝置100可以至少包括一儲存單元110與一處理單元120。Figure 1 shows a subjective opinion management apparatus for an article in accordance with an embodiment of the present invention. As shown in FIG. 1, the subjective opinion management apparatus 100 according to the article of the embodiment of the present invention can be applied to an electronic device such as a computer. The subject opinion management device 100 of the article may include at least one storage unit 110 and one processing unit 120.
儲存單元110可以包括至少一文章語句111、一屬性詞庫112、與一距離模型113。值得注意的是，在一些實施例中，文章之主觀意見管理裝置100可以更包括一網路連接單元(第1圖未顯示)，用以連接至一網路，如有線網路、電信網路、與無線網路等。藉由網路連接單元文章之主觀意見管理裝置100可以具有一網路接取能力，以連接至網路以取得一網路文章。在一些實施例中，文章語句111可以係網路文章中之一語句。屬性詞庫112中可以記錄相應至少一目標關鍵字，如物件之欲評價之複數候選屬性詞。舉例來說，目標關鍵字可以係一特定型號的手機，且相應此特定型號的手機的屬性詞可以包括電池、外型、功能、速度等。值得注意的是，在一些實施例中，儲存單元110可以更包括一評價詞庫、一程度詞庫、與一否定詞庫。在一些實施例中，「知網」所發佈之中文情感分析用詞語集中之情感詞子類別詞語集和評價詞子類別詞語集可以做為評價詞庫的來源，而程度詞子類別詞語集可以做為程度詞庫的來源。在一些實施例中，王正豪2010年於相關文獻中所歸納之否定詞可以做為否定詞庫的來源。必須說明的是，前述詞庫來源僅為本案之例子，本案並不限定於此。值得注意的是，在一些實施例中，意見片語可以包括評價詞、程度詞與否定詞。其中，前述評價詞庫、程度詞庫、與否定詞庫可以用來搜尋文章語句中之意見片語，相關細節將於後進行說明。距離模型113可以係由相關訓練資料所訓練出之文章語句中屬性詞與意見片語，尤其是意見片語中之評價詞間距離的分類模型。處理單元120係用以執行依據本案之文章之主觀意見 管理方法，其細節將於後進行說明。The storage unit 110 may include at least one article sentence 111, an attribute vocabulary 112, and a distance model 113. It should be noted that, in some embodiments, the subjective opinion management apparatus 100 of the article may further include a network connection unit (not shown in FIG. 1) for connecting to a network, such as a wired network or a telecommunication network. , with wireless networks, etc. The subjective opinion management device 100 by the network connection unit article may have a network access capability to connect to the network to obtain a web article. In some embodiments, the article statement 111 can be one of the sentences in the web article. The attribute lexicon 112 may record at least one target keyword, such as a plurality of candidate attribute words of the object to be evaluated. For example, the target keyword can be a specific model of mobile phone, and the attribute words of the mobile phone corresponding to the specific model can include battery, appearance, function, speed, and the like. It should be noted that, in some embodiments, the storage unit 110 may further include an evaluation vocabulary, a degree vocabulary, and a negative vocabulary. In some embodiments, the sentiment word subcategory word set and the evaluation word subcategory word set in the Chinese sentiment analysis word group published by "HowNet" can be used as the source of the evaluation lexicon, and the degree vocabulary subcategory word set can be As the source of the degree lexicon. In some embodiments, the negative words summarized by Wang Zhenghao in the relevant literature in 2010 can be used as a source of negative lexicon. It must be noted that the source of the above-mentioned thesaurus is only an example of the present case, and the present case is not limited thereto. It should be noted that in some embodiments, the commentary may include an evaluation word, a degree word, and a negative word. The foregoing evaluation vocabulary, degree vocabulary, and negative vocabulary can be used to search for the opinions in the article sentence, and the relevant details will be described later. The distance model 113 may be a classification model of the attribute words and opinions in the article sentences trained by the relevant training materials, especially the distance between the evaluation words in the opinion words. The processing unit 120 is configured to perform subjective opinions of the article according to the present case. The management method, the details of which will be explained later.
第2圖顯示依據本發明實施例之距離模型訓練方法。Figure 2 shows a distance model training method in accordance with an embodiment of the present invention.
首先，如步驟S210，取得複數文章語句。值得注意的是，在一些實施例中，可以以網路爬蟲方式配合搜尋引擎來搜尋相應目標關鍵字之一定數量的網路文章。其中，距離模型係相應此目標關鍵字領域的訓練結果。同時，可以分析每一文章之相關關鍵資訊，如文章本文、人氣指數、發文時間等等。取得網路文章之後，可以將這些文章進行標示，以確定係文章作者本身進行意見評論的文章，以將新聞、廣告、轉錄等文章濾除。接著，將標示的文章進行語句層級的斷句，其中具有主觀意見評論的文章語句將會被取出。舉例來說，擷取出之文章語句可以包括「就一直讓我對智慧手機的功能與實用性讚譽有加」、「還有下方的黑線部分就是一直被嫌棄的天線設計」、「運行速度更快」、「iPhone 4是”最好的智能手機”」、「電池很不經用」等。如步驟S220，搜尋文章語句中之屬性詞與意見片語。其中，屬性詞庫中之候選屬性詞可以比對文章語句，從而找出文章語句中之屬性詞及屬性詞於文章語句中的位置。此外，依據評價詞庫、程度詞庫、與否定詞庫搜尋文章語句中之意見片語及意見片語於文章語句中的位置。其中，意見片語可以包括評價詞、程度詞與否定詞。舉例來說，「功能」、「天線」、「速度」、「iPhone 4」、「電池」可以係前述文章語句中之屬性詞。「讚譽有加」、「嫌棄」、「更快」、「最好」、「不經用」可以係前述文章語句中之意見片語。接著，如步驟S230，依據屬性詞於文章語句中的位置及意見片語於文章語句中的位置計算至少一特徵值，並如步驟S240，將特徵值輸入距離模型以進行訓練。必須說明的是，在一些實施例中，距離模型可以係一支援向量機(Support Vector Machine,SVM)，且特徵輸入至支援向量機之前可以先進行正規化轉換為特徵值。值得注意的是，在一些實施例中，輸入至支援向量機的特徵可以包括屬性詞與評價詞的距離，其中正規化後特徵值的數值可以係大於0。在一些實施例中，輸入至支援向量機的特徵可以包括評價詞情感正反面，其中正規化後特徵值的數值可以係1或-1。在一些實施例中，輸入至支援向量機的特徵可以係屬性詞位置相對於評價詞的前後，其中正規化後特徵值的數值可以係1或-1。在一些實施例中，輸入至支援向量機的特徵可以係有無程度詞，其中正規化後特徵值的數值可以係1或-1。 在一些實施例中，輸入至支援向量機的特徵可以係程度詞分級，其中正規化後特徵值的數值可以係0-6。在一些實施例中，輸入至支援向量機的特徵可以係程度詞位置相對於評價詞的前後，其中正規化後特徵值的數值可以係1、0或-1。在一些實施例中，輸入至支援向量機的特徵可以係評價詞與否定詞之距離，其中正規化後特徵值的數值可以係大於等於0。在一些實施例中，輸入至支援向量機的特徵可以係有無否定詞，其中正規化後特徵值的數值可以係1或-1。在一些實施例中，輸入至支援向量機的特徵可以係否定詞位置相對於評價詞的前後，其中正規化後特徵值的數值可以係1、0或-1。在一些實施例中，輸入至支援向量機的特徵可以係否定詞與評價詞之距離，其中正規化後特徵值的數值可以係大於等於0。值得注意的是，在一些實施例中，支援向量機可以係產生距離模型之工具，其中，屬性詞與評價詞的距離可以作為分類標籤，其他特徵可以做為索引，而特徵所相應之特徵值作為數值，以輸入至支援向量機來進行訓練以產生主觀意見語句的距離分類模式。必須說明的是，前述特徵及其特徵值皆為本案之例子，本案並不限定於此。任何足以依據文章語句中屬性詞與意見片語產生之特徵皆可應用至本案中。First, in step S210, a plurality of article sentences are obtained. It is worth noting that in some embodiments, a web crawler may be used in conjunction with a search engine to search for a certain number of web articles of the corresponding target keyword. Among them, the distance model corresponds to the training result of this target keyword field. At the same time, you can analyze the relevant key information of each article, such as the article article, popularity index, time of publication, and so on. Once you have a web article, you can mark these articles to identify articles that are authored by the article authors themselves to filter out articles such as news, ads, and transcriptions. Next, the marked article is sentenced at the sentence level, and the article sentence with subjective comments will be taken out. For example, the article sentence that can be taken out can include "I have always praised the function and usability of smart phones" and "the black line below is the antenna design that has been abandoned" and "More speed" Fast, "iPhone 4 is the best smartphone", "Battery is not used" and so on. In step S220, the attribute words and the opinion words in the article sentence are searched. Among them, the candidate attribute words in the attribute lexicon can compare the article sentences, thereby finding the position of the attribute words and attribute words in the article sentence in the article sentence. In addition, according to the evaluation lexicon, the degree lexicon, and the negative vocabulary, the position of the commentary and the commentary in the article sentence are searched for in the article sentence. Among them, the opinion words may include evaluation words, degree words and negative words. For example, "function", "antenna", "speed", "iPhone 4", "battery" can be attribute words in the above article. "Appreciation", "dislike", "faster", "best", "disuse" can be the commentary in the above article. Next, in step S230, at least one feature value is calculated according to the position of the attribute word in the article sentence and the position of the opinion phrase in the article sentence, and in step S240, the feature value is input into the distance model for training. It should be noted that in some embodiments, the distance model may be a Support Vector Machine (SVM), and the feature may be first normalized into a feature value before being input to the support vector machine. It should be noted that in some embodiments, the feature input to the support vector machine may include the distance between the attribute word and the evaluation word, wherein the value of the normalized feature value may be greater than zero. In some embodiments, the feature input to the support vector machine may include evaluating the front and back of the word emotion, wherein the value of the normalized feature value may be 1 or -1. In some embodiments, the feature input to the support vector machine may be before and after the attribute word position relative to the evaluation word, wherein the value of the normalized feature value may be 1 or -1. In some embodiments, the feature input to the support vector machine may be a degree word, wherein the value of the normalized feature value may be 1 or -1. In some embodiments, the features input to the support vector machine may be graded, wherein the values of the normalized feature values may be 0-6. In some embodiments, the feature input to the support vector machine may be before and after the degree word position relative to the evaluation word, wherein the value of the normalized feature value may be 1, 0 or -1. In some embodiments, the feature input to the support vector machine may be the distance between the evaluation word and the negative word, wherein the value of the normalized feature value may be greater than or equal to zero. In some embodiments, the feature input to the support vector machine may be associated with a negative word, wherein the value of the normalized feature value may be 1 or -1. In some embodiments, the feature input to the support vector machine may be before and after the negative word position relative to the evaluation word, wherein the value of the normalized feature value may be 1, 0 or -1. In some embodiments, the feature input to the support vector machine may be the distance between the negative word and the evaluation word, wherein the value of the normalized feature value may be greater than or equal to zero. It should be noted that in some embodiments, the support vector machine may be a tool for generating a distance model, wherein the distance between the attribute word and the evaluation word may be used as a classification label, and other features may be used as an index, and the feature value corresponding to the feature As a numerical value, the distance classification mode is performed by inputting to the support vector machine to generate a subjective opinion sentence. It should be noted that the foregoing features and their characteristic values are examples of the present case, and the present invention is not limited thereto. Any feature that is sufficient to generate the attribute words and comments in the article sentence can be applied to the case.
第3圖顯示依據本發明實施例之文章之主觀意見管理方法。依據本發明實施例之文章之主觀意見管理方法適用於一電子裝置，如電腦。Figure 3 shows a subjective opinion management method for an article in accordance with an embodiment of the present invention. The subjective opinion management method of the article according to the embodiment of the present invention is applicable to an electronic device such as a computer.
首先，如步驟S310，取得一文章語句。值得注意的是，在一些實施例中，可以以網路爬蟲方式取得一網路文章，接著，將文章進行語句層級的斷句，以取出具有主觀意見評論的文章語句。如步驟S320，依據屬性詞庫搜尋文章語句中之一屬性詞。其中，屬性詞庫中之候選屬性詞可以比對文章語句，從而找出文章語句中之屬性詞及屬性詞於文章語句中的位置。如步驟S330，將屬性詞輸入距離模型，從而得到相應屬性詞之一標準距離值。提醒的是，距離模型可以係依據複數訓練語句產生，其中每一訓練語句至少包括一訓練屬性詞與一訓練意見片語。其中，訓練屬性詞與訓練意見片語間之一距離被作為至少一特徵值以輸入距離模型。另外，在一些實施例中，特徵可以包括評價詞情感正反面、屬性詞位置相對於評價詞的前後、有無程度詞、程度詞分級、程度詞位置相對於評價詞的前後、評價詞與否定詞之距離、有無否定詞、否定詞位置相對於評價詞的前後、 與/或否定詞與評價詞之距離等。其中，屬性詞與評價詞的距離可以作為分類標籤，其他特徵可以做為索引，而特徵所相應之特徵值作為數值，以輸入至支援向量機來進行訓練以產生主觀意見語句的距離分類模式。類似地，前述特徵及其特徵值皆為本案之例子，本案並不限定於此。任何足以依據文章語句中屬性詞與意見片語產生之特徵皆可應用至本案中。如步驟S340，由文章語句取得一意見片語。值得注意的是，在一些實施例中，可以依據評價詞庫、程度詞庫、與否定詞庫搜尋文章語句中之意見片語及意見片語於文章語句中的位置。注意的是，在一些實施例中，可以依據步驟S330得到之標準距離值來搜尋文章語句中的意見片語，以增加搜尋之效率。第4圖顯示依據本發明實施例之意見片語搜尋方法。如步驟S410，依據屬性詞於文章語句中之位置及標準距離值決定由文章語句搜尋意見片語之一搜尋起始位置。如步驟S420，依據此搜尋起始位置由文章語句中搜尋意見片語。類似地，如前所述，可以依據評價詞庫、程度詞庫、與否定詞庫搜尋文章語句中之意見片語。接著，如步驟S350，依據屬性詞於文章語句中之位置與意見片語於文章語句中之位置計算屬性詞與意見片語於文章語句中之一實際距離值。之後，如步驟S360，依據標準距離值與實際距離值判斷文章語句是否為一非主觀意見語句。注意的是，在一些實施例中，當實際距離值大於或小於標準距離值時，可以判定此文章語句為一非主觀意見語句。值得注意的是，在一些實施例中，判斷文章語句是否為一非主觀意見語句時可以設定一容許誤差值。當實際距離值大於標準距離值加一容許誤差值或小於標準距離值減一容許誤差值時，可以判定文章語句為一非主觀意見語句。First, in step S310, an article sentence is obtained. It is worth noting that, in some embodiments, a web article can be obtained in a web crawling manner, and then the article is sentenced at a sentence level to extract an article sentence having subjective opinion comments. In step S320, one of the attribute words in the article sentence is searched according to the attribute lexicon. Among them, the candidate attribute words in the attribute lexicon can compare the article sentences, thereby finding the position of the attribute words and attribute words in the article sentence in the article sentence. In step S330, the attribute word is input into the distance model, thereby obtaining a standard distance value of one of the corresponding attribute words. It is reminded that the distance model may be generated according to a plurality of training sentences, wherein each training sentence includes at least one training attribute word and one training opinion phrase. Wherein, a distance between the training attribute word and the training opinion piece is taken as at least one feature value to input the distance model. In addition, in some embodiments, the feature may include evaluating the front and back of the word emotion, the position of the attribute word relative to the evaluation word, the degree of presence or absence of the word, the degree of the degree word, the position of the degree word relative to the evaluation word, the evaluation word and the negative word. The distance, the presence or absence of a negative word, the position of the negative word relative to the evaluation word, And/or the distance between the negative word and the evaluation word. The distance between the attribute word and the evaluation word may be used as a classification label, and other features may be used as an index, and the feature value corresponding to the feature is used as a numerical value, and is input to the support vector machine for training to generate a distance classification mode of the subjective opinion statement. Similarly, the foregoing features and their characteristic values are examples of the present case, and the present invention is not limited thereto. Any feature that is sufficient to generate the attribute words and comments in the article sentence can be applied to the case. In step S340, an opinion phrase is obtained from the article sentence. It should be noted that, in some embodiments, the evaluation vocabulary, the degree lexicon, and the negative vocabulary may be used to search for the position of the commentary and the commentary in the article sentence in the article sentence. It is noted that, in some embodiments, the comment words in the article sentence may be searched according to the standard distance value obtained in step S330 to increase the efficiency of the search. Figure 4 shows a commentary phrase search method in accordance with an embodiment of the present invention. In step S410, a search start position of one of the article sentences is determined according to the position of the attribute word in the article sentence and the standard distance value. In step S420, a search phrase is searched for in the article sentence according to the search start position. Similarly, as described above, the commentary in the article sentence can be searched according to the evaluation lexicon, the degree lexicon, and the negative vocabulary. Next, in step S350, an actual distance value of the attribute word and the opinion phrase in the article sentence is calculated according to the position of the attribute word in the article sentence and the position of the opinion phrase in the article sentence. Then, in step S360, it is determined whether the article sentence is a non-subjective opinion statement based on the standard distance value and the actual distance value. It is noted that in some embodiments, when the actual distance value is greater than or less than the standard distance value, the article statement can be determined to be a non-subjective opinion statement. It should be noted that in some embodiments, an allowable error value may be set when determining whether the article statement is a non-subjective opinion statement. When the actual distance value is greater than the standard distance value plus an allowable error value or less than the standard distance value minus one allowable error value, the article sentence may be determined to be a non-subjective opinion statement.
第5圖顯示依據本發明另一實施例之文章之主觀意見管理方法。依據本發明實施例之文章之主觀意見管理方法適用於一電子裝置，如電腦。Figure 5 is a diagram showing a subjective opinion management method for an article according to another embodiment of the present invention. The subjective opinion management method of the article according to the embodiment of the present invention is applicable to an electronic device such as a computer.
首先，如步驟S510，取得一文章語句。類似地，在一些實施例中，可以以網路爬蟲方式取得一網路文章，接著，將文章進行語句層級的斷句，以取出具有主觀意見評論的文章語句。如步驟S520，依據屬性詞庫搜尋文章語句中之一屬性詞。其中，屬性詞庫中之候選屬性詞可以比對文章語句，從而找出文章語句中之屬性詞及屬性詞於文章語句中的位 置。如步驟S530，將屬性詞輸入距離模型，從而得到相應屬性詞之一標準距離值。類似地，距離模型可以係依據複數訓練語句產生，其中每一訓練語句至少包括一訓練屬性詞與一訓練意見片語。其中，訓練屬性詞與訓練意見片語間之一距離被作為至少一特徵值以輸入距離模型。另外，在一些實施例中，特徵可以包括評價詞情感正反面、屬性詞位置相對於評價詞的前後、有無程度詞、程度詞分級、程度詞位置相對於評價詞的前後、評價詞與否定詞之距離、有無否定詞、否定詞位置相對於評價詞的前後、與/或否定詞與評價詞之距離等。其中，屬性詞與評價詞的距離可以作為分類標籤，其他特徵可以做為索引，而特徵所相應之特徵值作為數值，以輸入至支援向量機來進行訓練以產生主觀意見語句的距離分類模式。類似地，前述特徵及其特徵值皆為本案之例子，本案並不限定於此。任何足以依據文章語句中屬性詞與意見片語產生之特徵皆可應用至本案中。如步驟S540，依據屬性詞於文章語句中之位置及標準距離值決定由文章語句搜尋意見片語之一搜尋起始位置。如步驟S550，依據此搜尋起始位置由文章語句中搜尋意見片語。類似地，如前所述，可以依據評價詞庫、程度詞庫、與否定詞庫搜尋文章語句中之意見片語。First, in step S510, an article sentence is obtained. Similarly, in some embodiments, a web article can be obtained in a web crawler manner, and then the article is sentenced at a sentence level to extract an article sentence having subjective opinion comments. In step S520, one of the attribute words in the article sentence is searched according to the attribute vocabulary. Among them, the candidate attribute words in the attribute lexicon can compare the article sentences, thereby finding the position of the attribute words and attribute words in the article sentence in the article sentence. Set. In step S530, the attribute word is input into the distance model, thereby obtaining a standard distance value of one of the corresponding attribute words. Similarly, the distance model may be generated in accordance with a plurality of training sentences, wherein each training statement includes at least one training attribute word and one training opinion phrase. Wherein, a distance between the training attribute word and the training opinion piece is taken as at least one feature value to input the distance model. In addition, in some embodiments, the feature may include evaluating the front and back of the word emotion, the position of the attribute word relative to the evaluation word, the degree of presence or absence of the word, the degree of the degree word, the position of the degree word relative to the evaluation word, the evaluation word and the negative word. The distance, the presence or absence of a negative word, the position of the negative word relative to the evaluation word, and/or the distance between the negative word and the evaluation word. The distance between the attribute word and the evaluation word may be used as a classification label, and other features may be used as an index, and the feature value corresponding to the feature is used as a numerical value, and is input to the support vector machine for training to generate a distance classification mode of the subjective opinion statement. Similarly, the foregoing features and their characteristic values are examples of the present case, and the present invention is not limited thereto. Any feature that is sufficient to generate the attribute words and comments in the article sentence can be applied to the case. In step S540, a search start position of one of the article comments is determined according to the position of the attribute word in the article sentence and the standard distance value. In step S550, the search phrase is searched for by the article sentence according to the search start position. Similarly, as described above, the commentary in the article sentence can be searched according to the evaluation lexicon, the degree lexicon, and the negative vocabulary.
因此，透過本案之文章之主觀意見管理方法及其相關裝置可以對於網路文章進行主觀意見之識別作業，從而增加由文章中擷取主觀意見之正確性，且/或增加由文章中搜尋意見片語之效率，並減少電子裝置之系統資源浪費。Therefore, the subjective opinion management method and related devices of the article in this case can identify the subjective opinions of the online articles, thereby increasing the correctness of the subjective opinions extracted from the articles, and/or increasing the search for comments from the articles. Language efficiency and reduce system resources waste of electronic devices.
本發明之方法，或特定型態或其部份，可以以程式碼的型態存在。程式碼可以包含於實體媒體，如軟碟、光碟片、硬碟、或是任何其他機器可讀取(如電腦可讀取)儲存媒體，亦或不限於外在形式之電腦程式產品，其中，當程式碼被機器，如電腦載入且執行時，此機器變成用以參與本發明之裝置。程式碼也可以透過一些傳送媒體，如電線或電纜、光纖、或是任何傳輸型態進行傳送，其中，當程式碼被機器，如電腦接收、載入且執行時，此機器變成用以參與本發明之裝置。當在一般用途處理單元實作時，程式碼結合處理單元提供一操作類似於應用特定邏輯電路之獨特裝置。The method of the invention, or a particular type or portion thereof, may exist in the form of a code. The code may be included in a physical medium such as a floppy disk, a CD, a hard disk, or any other machine readable (such as computer readable) storage medium, or is not limited to an external computer program product, wherein When the code is loaded and executed by a machine, such as a computer, the machine becomes a device for participating in the present invention. The code can also be transmitted via some transmission medium, such as a wire or cable, fiber optics, or any transmission type, where the machine becomes part of the program when it is received, loaded, and executed by a machine, such as a computer. Invented device. When implemented in a general purpose processing unit, the code combination processing unit provides a unique means of operation similar to application specific logic.
雖然本發明已以較佳實施例揭露如上，然其並非用以限定 本發明，任何熟悉此項技藝者，在不脫離本發明之精神和範圍內，當可做些許更動與潤飾，因此本發明之保護範圍當視後附之申請專利範圍所界定者為準。Although the invention has been disclosed above in the preferred embodiments, it is not intended to be limiting In the present invention, the scope of the present invention is defined by the scope of the appended claims, unless otherwise claimed.
S310、S320、…、S360‧‧‧步驟S310, S320, ..., S360‧‧ steps
Claims (18)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW102118215A TWI493364B (en) | 2013-05-23 | 2013-05-23 | Management methods for subjective comments of articles, and related devices and computer program products |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW102118215A TWI493364B (en) | 2013-05-23 | 2013-05-23 | Management methods for subjective comments of articles, and related devices and computer program products |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| TW201445335A TW201445335A (en) | 2014-12-01 |
| TWI493364B true TWI493364B (en) | 2015-07-21 |
Family
ID=52707057
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| TW102118215A TWI493364B (en) | 2013-05-23 | 2013-05-23 | Management methods for subjective comments of articles, and related devices and computer program products |
Country Status (1)
| Country | Link |
|---|---|
| TW (1) | TWI493364B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110738046B (en) * | 2018-07-03 | 2023-06-06 | 百度在线网络技术（北京）有限公司 | Viewpoint extraction method and apparatus |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200630825A (en) * | 2005-02-22 | 2006-09-01 | Webgenie Information Ltd | Method to automatically summarize chinese digital documents |
| US20090265238A1 (en) * | 2008-04-22 | 2009-10-22 | Jeong Hoon Lee | Method and system for providing content |
| TW201118780A (en) * | 2009-11-24 | 2011-06-01 | Univ Nat Chiao Tung | Intelligent mobile dervice product evaluation system and method based on information retrieval technnology |
-
2013
- 2013-05-23 TW TW102118215A patent/TWI493364B/en not_active IP Right Cessation
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200630825A (en) * | 2005-02-22 | 2006-09-01 | Webgenie Information Ltd | Method to automatically summarize chinese digital documents |
| US20090265238A1 (en) * | 2008-04-22 | 2009-10-22 | Jeong Hoon Lee | Method and system for providing content |
| TW201118780A (en) * | 2009-11-24 | 2011-06-01 | Univ Nat Chiao Tung | Intelligent mobile dervice product evaluation system and method based on information retrieval technnology |
Non-Patent Citations (1)
| Title |
|---|
| 蕭瑞祥,姜青山,曹金豐,簡之文,"部落格文章情感分析之研究," 2012年資訊管理國際研討會(ICIM 2012),2012/05/19 * |
Also Published As
| Publication number | Publication date |
|---|---|
| TW201445335A (en) | 2014-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109145153B (en) | Intention category identification method and device | |
| US9424524B2 (en) | Extracting facts from unstructured text | |
| Musto et al. | A Comparison of Lexicon-based Approaches for Sentiment Analysis of Microblog Posts. | |
| US20180189857A1 (en) | Recommendation through conversational ai | |
| Montejo-Ráez et al. | Ranked wordnet graph for sentiment polarity classification in twitter | |
| Bergsma et al. | Stylometric analysis of scientific articles | |
| Sidorov et al. | Syntactic n-grams as machine learning features for natural language processing | |
| Katz et al. | ConSent: Context-based sentiment analysis | |
| Hogenboom et al. | Multi-lingual support for lexicon-based sentiment analysis guided by semantics | |
| US9910930B2 (en) | Scalable user intent mining using a multimodal restricted boltzmann machine | |
| AU2014204091B2 (en) | Determining product categories by mining chat transcripts | |
| US20160203498A1 (en) | System and method for identifying and scoring leads from social media | |
| Shoukry et al. | A hybrid approach for sentiment classification of Egyptian dialect tweets | |
| US10037367B2 (en) | Modeling actions, consequences and goal achievement from social media and other digital traces | |
| CN111046221A (en) | Song recommendation method and device, terminal equipment and storage medium | |
| US11113722B2 (en) | Providing content related to sentiment of product feature | |
| US11599586B2 (en) | Method and system for providing alternative result for an online search previously with no result | |
| García-Moya et al. | Storing and analysing voice of the market data in the corporate data warehouse | |
| US20160350278A1 (en) | Claim polarity identification | |
| Nguyen et al. | Statistical approach for figurative sentiment analysis on social networking services: a case study on twitter | |
| CN105786971B (en) | A kind of grammer point recognition methods towards international Chinese teaching | |
| CN107665442B (en) | Method and device for acquiring target user | |
| TWI493364B (en) | Management methods for subjective comments of articles, and related devices and computer program products | |
| Mechti et al. | Machine learning for classifying authors of anonymous tweets, blogs, reviews and social media | |
| Lincy et al. | A survey on challenges in sentiment analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MM4A | Annulment or lapse of patent due to non-payment of fees |