{ "paper_id": "2019", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T07:30:11.011452Z" }, "title": "Dataset for Aspect Detection on mobile reviews in Hindi", "authors": [ { "first": "Ayush", "middle": [ "Joshi" ], "last": "Pruthwik", "suffix": "", "affiliation": { "laboratory": "", "institution": "IIIT Hyderabad", "location": {} }, "email": "pruthwik.mishra@research." }, { "first": "Mishra", "middle": [], "last": "Dipti", "suffix": "", "affiliation": { "laboratory": "", "institution": "IIIT Hyderabad", "location": {} }, "email": "dipti@iiit.ac.in" }, { "first": "Misra", "middle": [], "last": "Sharma", "suffix": "", "affiliation": { "laboratory": "", "institution": "IIIT Hyderabad", "location": {} }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "In recent years Opinion Mining has become one of the very interesting fields of Language Processing. To extract the gist of a sentence in a shorter and efficient manner is what opinion mining provides. In this paper we focus on detecting aspects for a particular domain. While relevant research work has been done in aspect detection in resource rich languages like English, we are trying to do the same in a relatively resource poor Hindi language. Here we present a corpus of mobile reviews which are labelled with carefully curated aspects. The motivation behind Aspect detection is to get information on a finer level about the data. In this paper we identify all aspects related to the gadget which are present on the reviews given online on various websites. We also propose baseline models to detect aspects in Hindi text after conducting various experiments.", "pdf_parse": { "paper_id": "2019", "_pdf_hash": "", "abstract": [ { "text": "In recent years Opinion Mining has become one of the very interesting fields of Language Processing. To extract the gist of a sentence in a shorter and efficient manner is what opinion mining provides. In this paper we focus on detecting aspects for a particular domain. While relevant research work has been done in aspect detection in resource rich languages like English, we are trying to do the same in a relatively resource poor Hindi language. Here we present a corpus of mobile reviews which are labelled with carefully curated aspects. The motivation behind Aspect detection is to get information on a finer level about the data. In this paper we identify all aspects related to the gadget which are present on the reviews given online on various websites. We also propose baseline models to detect aspects in Hindi text after conducting various experiments.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "Over the last decade people tend to search for products online rather than physically on stores. This has resulted in a surge of online forums where reviews are available on various products, electronic gadgets being one of the more popular ones. But reading so many long reviews is very time consuming and there is no uniformity on the parameters of reviews. To solve this, research work has been done in this area in the form of Aspect Detection which helps to point out the key specifications of the product in a structured format. But the work is limited to only worldwide languages as English and French. For a multi-lingual country like India, we are still far away in getting these information in the native language. We aimed at creating a dataset in Hindi which has the highest number of native speakers. The dataset is annotated with aspects for mobile reviews.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "An aspect is a word in a sentence which has some polarity associated with it. The aspect should hold major meaning of the sentence. Following examples will state what aspect is: S1 : \u0936\u093e\u0913\u092e\u0940 \u0930\u0947 \u0921\u092e\u0940 4\u090f \u0915\u094b \u092a\u0939\u0932\u0940 \u092c\u093e\u0930 \u0939\u093e\u0925 \u092e\u0947\u0902 \u0932\u0947 \u0928\u0947 \u092a\u0930 \u092f\u0939 \u0906\u092a\u0915\u094b \u092e\u0947 \u091f\u0932 \u092c\u0949\u0921\u0940 \u0915\u093e \u092c\u0928\u093e \u0932\u0917\u0947 \u0917\u093e \u0964 S1 : Xiaomi redmi 4A ko pehli baar hath m lene par yeh aapko metal body ka bana lagega.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Aspect1 : \"\u092e\u0947 \u091f\u0932 \u092c\u0949\u0921\u0940\" (metal body) which falls under the \"\u093f\u0921\u095b\u093e\u0907\u0928\" (design) category. The aspect shows importance by indicating how the mobile is built. S2: \u0936\u093e\u0913\u092e\u0940 \u0930\u0947 \u0921\u092e\u0940 \u0928\u094b\u091f \u092e\u0947\u0902 2 \u0917\u0940\u0917\u093e\u0939\u091f\u094d\u091c\u0930\u094d \u093c \u0911\u0915\u094d\u091f\u093e -\u0915\u094b\u0930 \u0915\u094d\u0935\u093e\u0932\u0915\u0949\u092e \u0938\u094d\u0928\u0948 \u092a\u0921\u0930\u094d \u0948 \u0917\u0928 625 \u092a\u0930\u094d\u094b\u0938\u0947 \u0938\u0930 \u0915\u093e \u0907\u0938\u094d\u0924\u0947 \u092e\u093e\u0932 \u0939\u0941 \u0906 \u0939\u0948 \u0964 S2: Xiaomi Redmi note m 2 gigahertz octacore qualcomm snapdragon 625 processor ka istemal hua hai. Aspect:\"\u0911\u0915\u094d\u091f\u093e -\u0915\u094b\u0930 \u0915\u094d\u0935\u093e\u0932\u0915\u0949\u092e \u0938\u094d\u0928\u0948 \u092a\u0921\u0930\u094d \u0948 \u0917\u0928 625\" (Octa core qualcomm snapdragon) which falls under the \"\u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928\" (specification) category . The aspect tells specifically tells the details of product. S3: \u0905\u092b\u0938\u094b\u0938 \u092f\u0939 \u093f\u0915 \u0906\u092a \u0909\u0928\u094d\u0939\u0947\u0902 \u0939\u091f\u093e \u0928\u0939\u0940\u0902 \u0938\u0915\u0924\u0947 \u0964 S3: Afsos yeh ki aap unhe hata nahi sakte. Aspect : \"NULL\" as there is no word which tells about any detail of the product. Hence, it is classified under no aspect category.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Major work has been done in Aspect Detection when it comes to resource rich languages like English. The work of Aspect Detection has also been followed by Sentiments analysis which plays a major part in Opinion mining. In 2014 SemEval-Task 4, Maria Pontiki (2014) provided the first dataset which con- sisted of English reviews annotated at sentence level with their aspects followed by their polarity. Some of the systems that emerged who targeted this task were Zhiqiang Toh (2014), Chernyshevich (2014); Joachim Wagner and Tounsi (2014); Giuseppe Castellucci (2014), Shweta Yadav (2015). However, almost all these systems are related to some specific languages, especially English. In 2016, SemEval released new datasets of similar domains(mobile, laptop, restaurant) 1 but in multiple languages. In 2016, the datasets were released in English, Arabic, Chinese, Dutch, French, Russian, Spanish and Turkish. But this area of field is largely unexplored in Indian languages due to the unavailability of high quality datasets and other tools and resources required. The datasets which were created by research groups mainly by Aditya Joshi (2010); Balamurali A R (2011, 2012) were very less in size and low in quality. Also Google transolator was used to create data in Indian languages (Akshat Bakliwal, 2012) but dataset created was not rich enough to perform aspect detection with high efficiency. Moreover the datasets available in Hindi were not domain specific which also added to poor results in past.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Related Work", "sec_num": "2" }, { "text": "As mentioned, earlier our work is on a specific domain. To build our corpus we scrapped data from various online forums with reviews on mobile phones. We extracted the text from the HTML data with the help of Beautiful-Soup library 2 in python. As our language was Hindi, online reviews were very less for which we tried both dynamic and manual crawling of data.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Data Creation", "sec_num": "3" }, { "text": "After crawling over 8 websites, we were able to get over 381 reviews. We retrieved 294 mobile reviews(37410 sentences) in a HTML format after extensive removal of noisy reviews. We had 294 HTML files which had raw data between different HTML tags. There was no uniformity in the reviews, even after extraction and tokenization of these reviews, Many reviews had proper headings like specifications, performance, price, design under which two-three paragraphs of text was present. But there were many reviews without any headings. To make it uniform and bring it to sentence level rather than paragraph level, we assigned the heading as labels to every sentence appearing under that heading in the review. This was our first annotation strategy. While assigning heading as aspects, there were certain sentences which had no heading above them. Such sentences were labelled as NULL. After this initial annotation, we had 18 classes of aspects in total. After doing analysis on our 18 classes, we observed a lot of overlapping between different classes. Some classes had the same name, but due to spelling variations they were assigned different labels. Table 3 gives a clear picture about the overlapping between different classes. We show the counts of highly frequent overlapping class pairs.", "cite_spans": [], "ref_spans": [ { "start": 1151, "end": 1158, "text": "Table 3", "ref_id": "TABREF5" } ], "eq_spans": [], "section": "Data Creation", "sec_num": "3" }, { "text": "Over \u2022 NULL and \u0939\u092e\u093e\u0930\u093e \u092b\u0948 \u0938\u0932\u093e(hamara faisla) were merged as into a single class NULL.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Class1 and Class2", "sec_num": null }, { "text": "After eliminating all these redundancies, we finally had 5 classes or aspects for our mobile reviews. Two annotators were involved in this task. We obtained a Fleiss' 3 score of 0.87 for inter annotator agreement.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Class1 and Class2", "sec_num": null }, { "text": "The main task was to predict aspects in every sentence in a review. We used different classifiers for the prediction task. We mostly experimented with machine learning models with 5-fold cross-validation as we had limited amount of data at our disposal.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimental Setup", "sec_num": "4" }, { "text": "Feature engineering is critical in designing accurate models. The features used in designing our supervised learning models are detailed here.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Feature Engineering", "sec_num": "4.1" }, { "text": "\u2022 Word n-grams -This feature deals with the presence or absence of certain sequence of words. The value of n used varied from 1 to 2.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "TF-IDF Vectors", "sec_num": null }, { "text": "\u2022 Character n-grams -This is similar to word n-grams where a sequence of characters is extracted from the text. The value of n used varied from 2 to 5.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "TF-IDF Vectors", "sec_num": null }, { "text": "We created baseline with two classifiers", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Machine Learning Approach", "sec_num": "4.2" }, { "text": "\u2022 Support Vector Machines (SVM)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Machine Learning Approach", "sec_num": "4.2" }, { "text": "\u2022 Multinomial Naive Bayes (MNB)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Machine Learning Approach", "sec_num": "4.2" }, { "text": "These two classifiers were implemented using the sklearn (Pedregosa et al., 2011) library. We used different feature set in both the classifiers.", "cite_spans": [ { "start": 57, "end": 81, "text": "(Pedregosa et al., 2011)", "ref_id": "BIBREF9" } ], "ref_spans": [], "eq_spans": [], "section": "Machine Learning Approach", "sec_num": "4.2" }, { "text": "The results are shown in table 5. Classifiers and their corresponding features are detailed in this table. We used precision, recall and macro F1-score as the evaluation metric for checking the performance of our models. The words 'uni', 'bi' refer to the word unigrams and bigrams respectively. char 'a-b' gram denotes the combination of character n-grams where n lies in {a, a + 1, a + 2, .., b}", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Results", "sec_num": "5" }, { "text": "From table 5, we observed that both the classifiers equally perform well on the data. We also observed that character n-grams models are superior than word n-gram models. Combination of word and char n-gram TF-IDF vectors do not significantly improve the performance.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Observation", "sec_num": "6" }, { "text": "From the values of confusion matrix, we observed that class \u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928(specification) has overshadowed classes NULL and \u0915\u0948 \u092e\u0930\u093e(camera). It shows that our model is not able to predict between the umbrella class and the child class accurately.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Observation", "sec_num": "6" }, { "text": "We annotated aspects for mobile reviews written in Hindi as a part of this work. We also presented baseline models for automatic aspect identification in mobile reviews. The baseline models will help us to annotate more reviews semi-automatically and can then be integrated to improve our systems. We will explore more into neural network architecture and word embeddings. The next task in this area would be to annotate polarity of the aspects. We can also explore identifying the most informative reviews (Mishra et al., 2017) .", "cite_spans": [ { "start": 507, "end": 528, "text": "(Mishra et al., 2017)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "Conclusion and Future Work", "sec_num": "7" }, { "text": "http://alt.qcri.org/semeval2016/task5/ index.php?id=data-and-tools", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "https://pypi.org/project/beautifulsoup4/", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "https://en.wikipedia.org/wiki/Fleiss' _kappa", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "A fall-back strategy for sentiment analysis in hindi: a case study", "authors": [ { "first": "Pushpak", "middle": [], "last": "Bhattacharyya", "suffix": "" }, { "first": "Aditya", "middle": [], "last": "Joshi", "suffix": "" }, { "first": "A", "middle": [ "R" ], "last": "Balamurali", "suffix": "" } ], "year": 2010, "venue": "Proceedings of ICON 2010: 8th International Conference on Natural Language Processing", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Pushpak Bhattacharyya Aditya Joshi, Balamurali A R. 2010. A fall-back strategy for sentiment analysis in hindi: a case study. In Proceedings of ICON 2010: 8th International Conference on Natural Language Processing, Macmillan Pub- lishers, India.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Hindi subjective lexicon : A lexical resource for hindi polarity classification", "authors": [ { "first": "Akshat", "middle": [], "last": "Vasudeva Varma", "suffix": "" }, { "first": "Piyush", "middle": [], "last": "Bakliwal", "suffix": "" }, { "first": "", "middle": [], "last": "Arora", "suffix": "" } ], "year": 2012, "venue": "Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC)", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Vasudeva Varma Akshat Bakliwal, Piyush Arora. 2012. Hindi subjective lexicon : A lexical re- source for hindi polarity classification. In Pro- ceedings of the Eight International Conference on Language Resources and Evaluation (LREC).", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "Harnessing wordnet senses for supervised sentiment classification", "authors": [ { "first": "Aditya Joshi Pushpak Bhattacharyya", "middle": [], "last": "Balamurali", "suffix": "" }, { "first": "A", "middle": [ "R" ], "last": "", "suffix": "" } ], "year": 2011, "venue": "Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing", "volume": "", "issue": "", "pages": "1081--1091", "other_ids": {}, "num": null, "urls": [], "raw_text": "Aditya Joshi Pushpak Bhattacharyya Balamurali A R. 2011. Harnessing wordnet senses for su- pervised sentiment classification. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, Edinburgh, Scot- land, UK, July 27-31., pages 1081-1091.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Cross-lingual sentiment analysis for indian languages using linked wordnets", "authors": [ { "first": "Aditya Joshi Pushpak Bhattacharyya", "middle": [], "last": "Balamurali", "suffix": "" }, { "first": "A", "middle": [ "R" ], "last": "", "suffix": "" } ], "year": 2012, "venue": "Proceedings of COLING 2012: Posters, COLING 2012", "volume": "", "issue": "", "pages": "73--82", "other_ids": {}, "num": null, "urls": [], "raw_text": "Aditya Joshi Pushpak Bhattacharyya Balamurali A R. 2012. Cross-lingual sentiment analysis for indian languages using linked wordnets. In Pro- ceedings of COLING 2012: Posters, COLING 2012, Mumbai, December 2012., pages 73-82.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Ihs rd belarus: Cross-domain extraction of product features using conditional random fields", "authors": [ { "first": "Maryna", "middle": [], "last": "Chernyshevich", "suffix": "" } ], "year": 2014, "venue": "Proceedings of the 8th International Workshop on Semantic Evaluation", "volume": "", "issue": "", "pages": "309--313", "other_ids": {}, "num": null, "urls": [], "raw_text": "Maryna Chernyshevich. 2014. Ihs rd belarus: Cross-domain extraction of product features us- ing conditional random fields. In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014) ., pages 309-313.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Unitor: Aspect based sentiment analysis with structured learning", "authors": [ { "first": "Danilo Croce Roberto Basili Giuseppe", "middle": [], "last": "Castellucci", "suffix": "" }, { "first": "Simone", "middle": [], "last": "Filice", "suffix": "" } ], "year": 2014, "venue": "Proceedings of the 8th International Workshop on Semantic Evaluation", "volume": "", "issue": "", "pages": "761--767", "other_ids": {}, "num": null, "urls": [], "raw_text": "Danilo Croce Roberto Basili Giuseppe Castellucci, Simone Filice. 2014. Unitor: Aspect based senti- ment analysis with structured learning. In Pro- ceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), Dublin, Ireland ., pages 761-767.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "Dcu: Aspect-based polarity classification for semeval task 4", "authors": [], "year": 2014, "venue": "Proceedings of the 8th International Workshop on Semantic Evaluation", "volume": "", "issue": "", "pages": "223--229", "other_ids": {}, "num": null, "urls": [], "raw_text": "Santiago Cortes Utsab Barman Dasha Bogdanova Jennifer Foster Joachim Wagner, Piyush Arora and Lamia Tounsi. 2014. Dcu: Aspect-based po- larity classification for semeval task 4. In Pro- ceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), Dublin, Ireland ., pages 223-229.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Semeval-2014 task 4: Aspect based sentiment analysis", "authors": [], "year": 2014, "venue": "John Pavlopoulos Haris Papageorgiou Ion Androutsopoulos Suresh Manandhar Maria Pontiki, Dimitrios Galanis", "volume": "", "issue": "", "pages": "27--35", "other_ids": {}, "num": null, "urls": [], "raw_text": "John Pavlopoulos Haris Papageorgiou Ion An- droutsopoulos Suresh Manandhar Maria Pon- tiki, Dimitrios Galanis. 2014. Semeval-2014 task 4: Aspect based sentiment analysis. In Proceed- ings of the 8th International Workshop on Se- mantic Evaluation (SemEval 2014)., pages 27- 35.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Iiit-h at ijcnlp-2017 task 3: A bidirectional-lstm approach for review opinion diversification", "authors": [ { "first": "Pruthwik", "middle": [], "last": "Mishra", "suffix": "" }, { "first": "Prathyusha", "middle": [], "last": "Danda", "suffix": "" }, { "first": "Silpa", "middle": [], "last": "Kanneganti", "suffix": "" }, { "first": "Soujanya", "middle": [], "last": "Lanka", "suffix": "" } ], "year": 2017, "venue": "Proceedings of the IJCNLP 2017, Shared Tasks", "volume": "", "issue": "", "pages": "53--58", "other_ids": {}, "num": null, "urls": [], "raw_text": "Pruthwik Mishra, Prathyusha Danda, Silpa Kan- neganti, and Soujanya Lanka. 2017. Iiit-h at ijcnlp-2017 task 3: A bidirectional-lstm ap- proach for review opinion diversification. In Proceedings of the IJCNLP 2017, Shared Tasks, pages 53-58.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "Scikit-learn: Machine learning in Python", "authors": [ { "first": "Fabian", "middle": [], "last": "Pedregosa", "suffix": "" }, { "first": "Ga\u00ebl", "middle": [], "last": "Varoquaux", "suffix": "" }, { "first": "Alexandre", "middle": [], "last": "Gramfort", "suffix": "" }, { "first": "Vincent", "middle": [], "last": "Michel", "suffix": "" }, { "first": "Bertrand", "middle": [], "last": "Thirion", "suffix": "" }, { "first": "Olivier", "middle": [], "last": "Grisel", "suffix": "" }, { "first": "Mathieu", "middle": [], "last": "Blondel", "suffix": "" }, { "first": "Peter", "middle": [], "last": "Prettenhofer", "suffix": "" }, { "first": "Ron", "middle": [], "last": "Weiss", "suffix": "" }, { "first": "Vincent", "middle": [], "last": "Dubourg", "suffix": "" } ], "year": 2011, "venue": "Matthieu Brucher, Matthieu Perrot, and Edouard Duchesnay", "volume": "12", "issue": "", "pages": "2825--2830", "other_ids": {}, "num": null, "urls": [], "raw_text": "Fabian Pedregosa, Ga\u00ebl Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Pret- tenhofer, Ron Weiss, Vincent Dubourg, Jake Vanderplas, Alexandre Passos, David Courna- peau, Matthieu Brucher, Matthieu Perrot, and Edouard Duchesnay. 2011. Scikit-learn: Ma- chine learning in Python. Journal of Machine Learning Research, 12:2825-2830.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Feature selection for entity extraction from multiple biomedical corpora: A pso-based approach", "authors": [ { "first": "Sriparna", "middle": [], "last": "Saha Shweta Yadav", "suffix": "" }, { "first": "Asif", "middle": [], "last": "Ekbal", "suffix": "" } ], "year": 2015, "venue": "Natural Language Processing and Information Systems", "volume": "", "issue": "", "pages": "220--233", "other_ids": {}, "num": null, "urls": [], "raw_text": "Sriparna Saha Shweta Yadav, Asif Ekbal. 2015. Feature selection for entity extraction from mul- tiple biomedical corpora: A pso-based approach. In In Natural Language Processing and Informa- tion Systems, Springer., pages 220-233.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "Dlirec: Aspect term extraction and term polarity classification system", "authors": [ { "first": "Wenting", "middle": [], "last": "Wang", "suffix": "" }, { "first": "Zhiqiang", "middle": [], "last": "Toh", "suffix": "" } ], "year": 2014, "venue": "Proceedings of the 8th International Workshop on Semantic Evaluation", "volume": "", "issue": "", "pages": "235--240", "other_ids": {}, "num": null, "urls": [], "raw_text": "Wenting Wang Zhiqiang Toh. 2014. Dlirec: As- pect term extraction and term polarity classi- fication system. In Proceedings of the 8th In- ternational Workshop on Semantic Evaluation (SemEval 2014), Dublin, Ireland, August 23-24, pages 235-240.", "links": null } }, "ref_entries": { "TABREF0": { "content": "
\u0938\u0949\u092b\u094d\u091f\u0935\u0947 \u092f\u0930software52
\u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 \u0914\u0930 \u095e\u0940\u091a\u0930Specification aur feature360
\u0939\u092e\u093e\u0930\u093e \u095e\u0948 \u0938\u0932\u093ehamara faisla9
\u0915\u0948 \u092e\u0930\u093e \u0914\u0930 \u092c\u0948 \u091f\u0930\u0940 \u0932\u093e\u0907\u092bcamera aur battery life5
\u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 \u0914\u0930 \u0938\u0949\u092b\u094d\u091f\u0935\u0947 \u092f\u0930specification aur software137
\u0915\u0948 \u092e\u0930\u093ecamera76
\u092a\u0930\u092b\u0949\u092e\u0947 \u0930\u094d \u0902 \u0938 \u0932\u0941 \u0915 \u0935 \u092c\u0928\u093e\u0935\u091fperformance look vah banawat826 26
\u092c\u0948 \u091f\u0930\u0940 \u0932\u093e\u0907\u092bbattery life1
\u0939\u092e\u093e\u0930\u093e \u092b\u0948 \u0938\u0932\u093ehamara faisla300
\u0915\u0948 \u092e\u0930\u093e \u092a\u0930\u092b\u0949\u092e\u0947 \u0930\u094d \u0902 \u0938camera performance16
\u093f\u0921\u095b\u093e\u0907\u0928design138
\u093f\u0921\u095b\u093e\u0907\u0928 \u0914\u0930 \u0932\u0941 \u0915design aur look168
NULLNULL352
\u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928specification139
\u093f\u0921\u095b\u093e\u0907\u0928 \u0914\u0930 \u093f\u092c\u0932\u094d\u0921design aur build390
\u093f\u0921\u095b\u093e\u0907\u0928 \u0914\u0930 \u093f\u0921\u0938\u094d\u092a\u094d\u0932\u0947design aur display49
\u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 , \u0938\u0949\u092b\u094d\u091f\u0935\u0947 \u092f\u0930 \u0914\u0930 \u092a\u0930\u092b\u0949\u092e\u0947 \u0930\u094d \u0902 \u0938 specification, software aur performance40
", "text": "", "html": null, "type_str": "table", "num": null }, "TABREF1": { "content": "", "text": "", "html": null, "type_str": "table", "num": null }, "TABREF3": { "content": "
", "text": "", "html": null, "type_str": "table", "num": null }, "TABREF5": { "content": "
\u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928, \u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 , \u0938\u0949\u092b\u094d\u091f\u0935\u0947 \u092f\u0930 \u0914\u0930 \u092a\u0930\u092b\u0949\u092e\u0947 \u0930\u094d \u0902 \u0938 (specification, specification, soft-
ware aur perfomance) clubbed under one
single class called \u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 (specifica-
tion).
\u2022 \u0915\u0948 \u092e\u0930\u093e \u0914\u0930 \u092c\u0948 \u091f\u0930\u0940 \u0932\u093e\u0907\u092b (camera aur battery life), \u0915\u0948 \u092e\u0930\u093e (camera), , \u0915\u0948 \u092e\u0930\u093e \u092a\u0930\u092b\u0949\u092e\u0947 \u0930\u094d \u0902 \u0938 (cam-
era performance) were clubbed under a
class \u0915\u0948 \u092e\u0930\u093e (camera).
\u2022 \u0932\u0941 \u0915 \u0935 \u092c\u0928\u093e\u0935\u091f (look wh banawat), \u093f\u0921\u095b\u093e\u0907\u0928
(design), \u093f\u0921\u095b\u093e\u0907\u0928 \u0914\u0930 \u0932\u0941 \u0915 (design aur look),
\u093f\u0921\u095b\u093e\u0907\u0928 \u0914\u0930 \u093f\u092c\u0932\u094d\u0921 (design aur build), \u093f\u0921\u095b\u093e\u0907\u0928 \u0914\u0930 \u093f\u0921\u0938\u094d\u092a\u094d\u0932\u0947 (design aur display) categorized under one class \u093f\u0921\u095b\u093e\u0907\u0928(design).
\u2022 \u0915\u0948 \u092e\u0930\u093e (camera),\u0915\u0948 \u092e\u0930\u093e \u092a\u0930\u092b\u0949\u092e\u0947 \u0930\u094d \u0902 \u0938 (camera per-formance), \u0915\u0948 \u092e\u0930\u093e \u0914\u0930 \u092c\u0948 \u091f\u0930\u0940 \u0932\u093e\u0907\u092b(camera
aur battery life) were categorized under
one class \u0915\u0948 \u092e\u0930\u093e (camera).
", "text": "Overlapping Between Initial ClassesThe following decisions to club different classes and provide them a single label were taken based on the percentage of overlapping.\u2022 \u0938\u0949\u092b\u094d\u091f\u0935\u0947 \u092f\u0930(software), \u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 \u0914\u0930 \u095e\u0940\u091a\u0930 (specification aur feature), \u0938\u094d\u092a\u0947 \u093f\u0938\u093f\u092b\u0915\u0947 \u0936\u0928 \u0914\u0930 \u0938\u0949\u092b\u094d\u091f\u0935\u0947 \u092f\u0930 (specification aur software),", "html": null, "type_str": "table", "num": null }, "TABREF7": { "content": "", "text": "Classwise Distribution", "html": null, "type_str": "table", "num": null }, "TABREF9": { "content": "
", "text": "Results Of Models After 5-fold Cross Validation", "html": null, "type_str": "table", "num": null } } } }