240 likes | 481 Views
Mining feature-opinion pairs and their reliability scores from web opinion sources. A. Kamal, M. Abulaish, and T. Anwar International Conference on Web Intelligence, Mining and Semantics ( WIMS ) 2012. Presented by Sole. Introduction. Opinion Data: user-generated content. Forums.
E N D
Mining feature-opinion pairs and their reliability scores from web opinion sources A. Kamal, M. Abulaish, and T. Anwar International Conference on Web Intelligence, Mining and Semantics (WIMS) 2012 Presented by Sole
Introduction • Opinion Data: user-generated content Forums Discussion Groups Blogs • Opinion Sources Customer Manufacturer
Introduction • Problems with Reviews Information Overload Biased Information Time Consuming • Solution • Approach to • Extract feature-opinions pairs from reviews • Determine the reliability score of each pair
Related Work • Relatively new area of study • Information Retrieval • Classification of positive/negative reviews • NLP, text mining, probabilistic approaches • Identify patterns on text to extract attribute-value pairs
Proposed Approach • Architecture of the system
Pre-processing • Review crawler • Noisy reviews are removed • Eliminate reviews created with no purpose or to increase/decrease the popularity of a product • Markup language is filtered • Remaining content is divided into manageable sizes • Boundaries are determined based on heuristics, e.g., granularity of words, stemming, synonyms
Document Parser • Text Analysis • Assigns Part-Of-Speech (POS) tags to each word • Converts each sentence into a set of dependency relations between pairs of words • Facilitate information extraction Noun Phrases Adjectives Adverbs Product features Opinions Degree of expressiveness of opinions
Feature and Opinion Learner • Feature-opinion learner • Analyzes dependency relations generated by a document • Generates all the possible information components from the documents • Information component: <f, m, o> • f refers to feature • m refers to modifier • o refers to an opinion
Feature and Opinion Learner • Rule 1 • In a dependency relation R, if there exist relationships nn(w1,w2) and nsubj(w3,w1) such that POS(w1)=POS(w2)=NN*, POS(w3)=JJ* and w1,w2are not stop-words, or if there exists a relationship nsubj(w3,w4) such that POS(w3)=JJ*, POS(w4)=NN* and w3,w4are not stop-words, then either (w1,w2) or w4is considered as a feature and w3as an opinion.
Feature and Opinion Learner • Rule 2 • In a dependency relation R, if there exist relationships nn(w1,w2) and nsubj(w3,w1) such that POS(w1)=POS(w2)=NN*, POS(w3)=JJ* and w1,w2 are not stop-words, or if there exists a relationship nsubj(w3,w4) such that POS(w3)=JJ*, POS(w4)=NN* and w3,w4are not stop-words, then either (w1,w2) or w3is considered as a feature and w4as an opinion. Thereafter, the relationship advmod(w3,w5) relating w3 with some adverbial word w5is searched. In case of presence of advmod relationship, the information componentis identified as <(w1,w2) or w4,w5,w3> otherwise <(w1,w2) or w4,-,w3>.
Feature and Opinion Learner • Rule 3 • In a dependency relation R, if there exist relationships nn(w1,w2) and nsubj(w3,w1) such that POS(w1)=POS(w2)=NN*, POS(w3)=VB* and w1,w2are not stop-words, or if there exist a relationship nsubj(w3,w4) such that POS(w3)=VB*, POS(w4)=NN* and w4is not a stop-word, then we search for acomp(w3,w5) relation. If acomp relationship exists such that POS(w5)=JJ* and w5is not a stop-word then either (w1,w2) or w4is assumed as the featureand w5as an opinion. Thereafter, the modifieris searched and the information component is generated in the same way as in Rule-2.
Feature and Opinion Learner • Rule 4 • In a dependency relation R, if there exist relationships nn(w1,w2) and nsubj(w3,w1) such that POS(w1)=POS(w2)=NN*, POS(w3)=VB* and w1,w2are not stop-words, or if, there exist a relationship nsubj(w3,w4) such that POS(w3)=VB*, POS(w4)=NN* and w4is not a stop-word, then we search for dobj(w3,w5) relation. If dobjrelationship exists such that POS(w5)=NN* and w5is not a stop-word then either (w1,w2) or w4is assumed as the feature and w5as an opinion.
Feature and Opinion Learner • Rule 5 • In a dependency relation R, if there exists a amod(w1,w2) relation such that POS(w1)=NN*, POS(w2)=JJ*, and w1and w2are not stop-words then w2is assumed to be an opinion and w1as an feature.
Feature and Opinion Learner • Rule 6 • In a dependency relation R, if there exists relationships nn(w1,w2) and nsubj(w3,w1) such that POS(w1)=POS(w2)=NN*, POS(w3)=VB* and w1,w2are not stop-words, or if there exists a relationship nsubj(w3,w4) such that POS(w3)=VB*, POS(w4)=NN* and w4is not a stop-word, then we search for dobj(w3,w5) relation. If dobj relationship exists such that POS(w5)=NN* and w5is not a stop-word then either (w1,w2) or w4is assumed as the feature and w5as an opinion. Thereafter, the relationship amod(w5,w6) is searched. In case of presence of amod relationship, if POS(w6)=JJ* and w6is not a stop-word, then the information component is identified as <(w1,w2) or w4,w5,w6> otherwise <(w1,w2) or w4, w5,->.
Feature and Opinion Learner • Example • Consider the following opinion sentences related to Nokia N95 • The screen is very attractive and bright • The sound some times comes out very clear • Nokia N95 has a pretty screen • Yes, the push mail is the “Best” in the business
Reliability Score Generator • Reliability Score • Removes noise due to parsing errors • Addresses contradicting opinions in reviews
Reliability Score Generator • HITS Algorithm • A higher score value for a pair reflects a tight integrity of the 2 components in a pair • The Hubs and Authority scores are computed iteratively • The feature score is calculated using the term frequency and inverse sentence frequency in each sentence of the document Authority Based on feature score and opinion score Hub
Reliability Score Generator • Pseudocode 1G := set of pages 2for each page p in Gdo 3p.auth = 1 // p.auth is the authority score of the page p 4p.hub = 1 // p.hub is the hub score of the page p 5 function HubsAndAuthorities(G) 6for step from 1 to k do // run the algorithm for k steps 7for each page p in Gdo // update all authority values first 8p.auth = 0 9for each page q in p.incomingNeighborsdo // p.incomingNeighbors is the set of pages that link to p 10 p.auth += q.hub 11for each page p in Gdo // then update all hub values 12p.hub = 0 13 for each page r in p.outgoingNeighborsdo //p.outgoingNeighbors is the set of pages that p links to 14p.hub += r.auth
Experimental Results • Dataset 4333 noun (or verb) and adjective pairs 1366 candidate features obtained after filtering 400 review • Sample list of extracted features, opinions, modifiers
Experimental Results • Metrics • True positive (TP): number of feature-opinion pairs that the system identifies correctly • False positive (FP): number of feature-opinion pairs that are falsely identified by the system • False negative (FN): number of feature-opinion pairs that the system fails to identify
Experimental Results • Feature and Opinion Learner • Precision: 79.3% • Recall: 70.6% • F-Measure: 74.7% • Observations • Direct and strong relationships between noun and adjectives cause non-relevant feature-opinion pairs • Lack of grammatical correctness in reviews affects the results yielded by NLP parsers • Recall values lower than precision indicates the inability of the systems to extract certain feature-opinion pairs correctly
Experimental Results • Sample results for different products • Observations • Lack of variations on metric values indicates the applicability of the proposed approach regardless of the domain of the review documents
Experimental Results • Reliability Score Generator Top-5 hub scored feature and opinion pairs and their reliability scores Sample feature-opinion pairs along with their hub and reliability scores
Conclusions • Future Work • Refine rules to improve precision and identify implicit features • Handle informal text common in reviews Reviews + Rules Hits Algorithm Feature-pinion pairs Reliability scores