Information Extraction for Vietnamese Real-Estate Advertisements by Pham Vi Lien Faculty of Information Technology University of Engineering and Technology Vietnam National University, Hanoi Supervised by Dr. Pham Bao Son A thesis submitted in fulfillment of the requirements for the degree of Master of Information Technology June, 2012 LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com ORIGINALITY STATEMENT I hereby declare that this thesis is my own work and to the best of my knowl- edge, it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at University or any other educational institution, ex- cept where due to acknowledgment is made to the thesis. Any contribution made in the research by others, with whom I have worked at University of Engineering and Technology or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to conception or in style, presentation and linguistic expression are acknowledged.
Signed: Date: i LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com Abstract In recent years, real-estate market in Vietnam is growing rapidly which creates a lot of information about real-estate, especially information on advertising for buying and selling activities of real-estate development. This poses an essential demand for building an information extraction system to help users deal with the increasing amount of real-estate advertisements on the Internet. We propose a rule-based approach to build an information extraction system for online real- estate advertisements in Vietnamese. At the same time, we set up a process to build an annotated corpus which can be used in machine learning approaches at a later stage.
Our system achieve promising results with F-measures of above 90%. Our approach is particularly suitable for under-resourced languages where an annotated corpus of a decent size is not readily available. Keywords: natural language processing, information extraction, online real- estate advertisements LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com Acknowledgements I would like to express great gratitude to my supervisor, Dr. Pham Bao Son, in the Faculty of Information Technology at University of Engineering and Technology of Vietnam National University, Hanoi, for his encouragement, support, patience, guidance and advice.
Without his constant invaluable direction and tolerance, I could not have become a better researcher. I would also like to respect to my lecturers who has taught me educational sub- jects at University of Engineering and Technology of Vietnam National University, Hanoi. I would also like to depict my great pleasure to my sponsor, Quang Trung University, who have granted me the full scholarship to follow my Master degree. I owe all friends and colleagues a huge thank for their encouragement and friend-ship.
They have provided great mental support to me when I got stressful at times. Last but not least, thank to my wife for her sympathy and love during the past years. I heartily thank my parents, parents-in-law and my sisters for their encouragement and the many years of support during my studies. Again, I owe my success in life as I am today to my parent’ unconditional love, hard work, and sacrifices To all, I thank you.
iii LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com Contents ORIGINALITY STATEMENT i Abstract ii Acknowledgements iii List of Figures vi List of Tables vii 1 Introduction 1 1.1 Problem and Idea .2 Scope of the thesis .1 Rule-based approach .2 Machine-learning approach .2 General Architecture of GATE .3 An example: ANNIE - A Nearly-New Information Extrac- tion System .4 Working with GATE. 13 3 Our Vietnamese Real-Estate Information Extraction system 14 3.1 Criterion of data collection. 17 iv LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com Table of Contents v 3.1 Remove incorrect Lookup annotations .2 Recognizing <TypeEstate> entities .3 Recognizing <CategoryEstate> entities .4 Recognizing <Zone> entities .5 Recognizing <Area>, <Price> and <Telephone> entities .6 Recognizing <Fullname> entities .7 Recognizing <Address> entities .8 Recognizing <Email> entities. 34 4 Experiments and Error Analysis 35 4.
40 5 Conclusion and Future Works 42 5. 42 A A typical code 44 B Relevant Publications 46 Bibliography 47 LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com List of Figures 1.1 The result for query "cần mua nhà ở Hà Nội" on Google Search .2 The expected result of our system.1 A screenshot of a GUI in GATE framework .2 The general architecture of GATE .1 Template of our system .2 An example of an original news article before normalization .3 An example of a normalized news article .4 The process of creating an annotated corpus and system development 21 3.5 The main code is defined to create a new Callisto task.6 A news articles annotated by Callisto .7 Architecture of our Vietnamese Real-Estate Information Extraction system .8 Typical Vietnamese Real-Estate Information Extraction system com- ponents .1 The performance of our system in three versions.2 Using lenient criteria to evaluate the annotation in three versions.3 Using strict criteria to evaluate the annotation in three versions.1 The screenshot of Real-Estate Information Extraction system .1 A code recognize TypeEstate entity .2 A code recognize Telephone entity .3 A code recognize Email entity .4 A code recognize Zone entity. 45 vi LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com List of Tables 4.1 Performance on the T raining3 data using lenient criteria .2 Performance on the T raining3 data using strict criteria .3 Performance on the T est3 data using lenient criteria .4 Performance on the T est3 data using strict criteria. 38 vii LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com Chapter 1 Introduction As data and information sources are growing rapidly everyday, dealing with this data become a big and challenging problem.
Popular techniques such as machine learning can not be easily applied for many language processing tasks in Viet- namese due to the lack of annotated corpora. This is indeed the case for pro- cessing real-estate advertising information. In this thesis, we propose to build an information extraction system for real-estate adverstisements in Vietnamese.1 Problem and Idea With the advent and development of the Internet, a great amount of data has been posted to the Internet. Those data are not only text but also image, audio, video, and so on.
They appear in most areas of life from economic, politic, society, medicine to the emerging areas today such as securities, finance, real-estate, etc. The explosion of data is constantly increasing everyday, especially, in the cloud computing age. Almost all of user data is stored on the web platform. This huge data source contain a lot of information.
If data are increasing rapidly, it means that, information is also growing much faster than data. With more information, users become more confused because the useful in- formation that they need is drifting following the stream-data. In order to help 1 LUAN VAN CHAT LUONG download : add luanvanchat@agmail. Introduction 2 people deal with this situation, there are many search engines that have been cre- ated such as Google1 , Bing2 , Yahoo3 , etc.
They quickly become an indispensable tool to assist human in finding useful information from the huge data sources on the Internet. However, they still haven’t met the expectations of the users, espe- cially, in the case where the user’s query is a question. Take the following example: We use the phrase "cần mua nhà ở Hà Nội" (buy a house in Hanoi) as a query for Google’s search engine (Figure 1. The results which we obtained is a list of links.
These links refer to websites containing one of the words of the above query.1, we can easily see that these results aren’t the expected results of the users. Users have to spend a lot of time to find an answer for their query from this list of links. Therefore, our desire is that the users should get a list of specific answers to the query.1: The result for query "cần mua nhà ở Hà Nội" on Google Search In order to solve the above problem, the researchers have looked into areas such as information extraction, text summarization, data mining, etc. to deliver more useful and specific information to users.
Information Extraction is one of the important tasks in natural language pro- cessing. The main idea of an information extraction system is to extract snippets 1 https://www.com/ 2 https://www.com/ 3 https://www.com/ LUAN VAN CHAT LUONG download : add luanvanchat@agmail. Introduction 3 of information from unstructured or semi-structured documents to fill in a struc- tured form which is called a template. In other words, the system will extract the requisite information from the content of the input documents to fill in the defined template output.
The data obtained after the extraction process can be presented directly for users or used as input data for third party applications such as analysis and prediction, information retrieval, data mining or search engine, etc. Around thirty years ago, information extraction started its rapid development. There are many studies in many different domains that have been publised. Most information extraction systems of the first generation are research and experimen- tation for documents in English.
In recent years, many studies of this technology have gradually appeared in other languages such as French, Japanese, Chinese, etc. However, it is still a new problem in Vietnamese, especially in the domain for real-estate advertisements. Our thesis addresses the problem of information extraction for Vietnamese online real-estate advertisements. We propose a rule-based approach for building this system.
At the same time, we also build an annotated corpus for the same task. There are several other approaches that have been used to tackle the above problem such as machine learning and hybrid method. However, there aren’t any annotated corpus in Vietnamese publicized for the community, especially in the real-estate field. So our rule based approach for this problem is reasonable and appropriate at this moment.
We can reduce the labour cost compared to other approaches.2 shows and and input sample and its expected output for our system.2: The expected result of our system. LUAN VAN CHAT LUONG download : add luanvanchat@agmail.2 Scope of the thesis With the development of the Internet, online advertising is practical and increas- ingly popular. It is an effective advertising solution for both advertising individuals, agencies and viewers. Thus, the data source from the advertisements is extremely large and diverse.
Our thesis focuses on processing the free online Vietnamese text advertisement in the real-estate domain.3 Thesis’ structure Our thesis is organized into five chapters as follows: • Chapter 1: We introduce the problem and idea to build a system to extract information for online real-estate advertisements in Vietnamese. • Chapter 2: We present an overview of related research for information ex- traction methods in general and real-estate domain in particular. • Chapter 3: We describe in details of how to build our Vietnamese Real-Estate Information Extraction system. • Chapter 4: We present the results of our experiments and the analysis of some failures.
• Chapter 5: We conclude with discussion about future development for the system. LUAN VAN CHAT LUONG download : add luanvanchat@agmail.com Chapter 2 Related Work Information extraction is one of commonly researched tasks of natural language processing. This is a technique to extract relevant information from unstructured or semi-structured text and present it in a structured format. In most of the cases, the extraction tasks concern processing human language texts [1–5].
Recently, they have been also used for the image/audio/video data [6–8]. At the end of 19981 , the Message Understanding Conferences (MUC) pro- gramme had arrived at a definition of information extraction which is split into five subtasks: • Named Entity recognition (NE): Finds and classifies names, places, etc. • Coreference resolution (CO): Identifies identity relations between entities. • Template Element construction (TE): Adds descriptive information to NE results (using CO).
• Template Relation construction (TR): Finds relations between TE entities.