Searching real estate property is uncommon task for most of the users. As a result, the user is not familiar with the detailed search condition which is useful for search. In this paper, we propose to use voice recognition as a support for real estate property search. First user set vague search condition with GUI. Then system listens conversation between users. From the conversation, system extracts keyword and suggest detailed search condition and real estate information search results with those conditions. We will discuss system design, algorithm used to link spoken words and detailed search condition and preliminary test results.
Typical user can set the vague request to search the real estate property like the town they want to search, layout of property and price. Further, for example, it is also possible to set detailed search conditions such as "Pets allowed". However, ©2018. Copyright for the individual papers remains with the authors. Copying permitted for private and academic purposes.
WII’18, March 11, 2018, Tokyo, Japan in most cases the user does not know what type of detailed search conditions are available and how to set them. As a result, they are not used effectively in the real estate search process.
To solve this issue, we assumed that we need support system which can recommend proper detailed search condition to assist real estate search. To do that, we have chosen to utilize the conversation between users. Asking detailed demand for new real estate property is sometimes intimidating for the user. Moreover, the user may not necessary know the proper word to search the real state information. Therefore, we assumed that the system which can extract detailed search condition from the casual user conversation would be effective to support the real estate property search.
Interfaces for searching information using voice have been studied for a long time, and commercial service which utilizes voice interface become popular in recent years due to popularization of smartphones and home devices. Recently speech recognition on the server side has become widespread, the accuracy of speech recognition has dramatically improved even for unspecified speakers.
However, in most cases speech recognition is used for simple
web search and search in the app store[
Considering technical difficulty described above, we have
chosen to use voice recognition as a support role. The user is
not assumed to directly talk to the search interface. Instead,
system listens the conversation between users. The system
recognizes the keyword in the conversation, and tries to
explore the appropriate detailed search condition to search the
real estate property. Matching spoken word by the user and
detailed search condition is the first challenge in this research.
The frustration that the user feels when using the voice
interface is that the system can not understand the meaning of the
word they have pronounced. We need to match appropriate
detailed search condition to words in users’ conversation.
There are various attempt to incorporate speech interface in
search as a support role for exploratory search. Andolina
et al. proposed systems [
Other reason why an existing system using speech recognition
makes a user feel frustrated is that the system is not transparent
[
SYSTEM DESIGN Screen shot of developed system is shown in Figure 1. User can set the search conditions which are area to search, layout and price of real estate property via GUI. After setting the search condition, users have free conversation about their intention for new house between themselves. System continuously monitors their conversation and recognized text is shown in the lower part of the screen. System automatically and continuously extract keywords from conversation and tries to find the related detailed search condition such as "Within 800 meters from convenience store", "Pet allowed". System also display the search result of the real estate property specifying each detailed condition. Link is shown between detailed search condition and searched property. There are two major challenges in developing the proposed system. First, we needed to develop algorithm which search detailed search condition using user’s conversation data. Second, we need interaction interface which will effectively support user to search real estate information even if the system does not recognize the user’s intention correctly. We will discuss these challenges next.
MATCHING ALGORITHM BETWEEN DETAILED SEARCH CONDITION AND SPOKEN WORDS One of the major frustrations felt using existing speech interface is that the system recognizes only programmed keywords while there is no clue for the user about which word to speak. As a result, quite often system does not recognize the word that the user pronounced. Even though the user pronounces the word which has similar meaning to the keyword that system recognizes, system can not understand the similarity between those words.
Similar difficulty is recognized in the field of question and
answer retrieval(herein Q&A retrieval) task[
In this research, we tried to utilize Word2Vec [
Next we will discuss the training data set for Word2Vec.
Related words extracted using Word2Vec depends on nature of
training data set. As an example, we will show the result of
retrieved related words of the word "Noise" using Wikipedia[
We have also gathered text data from the web site called All
About Japan [
However, number of words in All About Japan is not necessary enough. As a result, in many cases we could not extract related words because spoken word does not exist in All About Japan data set. Considering characteristics of each data set described above, we decided to use both data set for keyword extraction. First we tries to extract keywords from All About Japan data set. If no matching word is found, we tries to extract keywords using Wikipedia data set.
USER INTERACTION By using the algorithm described above, we can expect that we will be able to search the detailed search condition better than simple word match. However, still there are errors and we don’t expect that we can search the detailed search condition with high precision.
The other aspect that we should consider is transparency of the system which is described earlier. To ease user frustration, we need to make inference process clear to the user. Considering these factors, we designed the interaction interface shown in Figure 1. Not only detailed search condition and spoken words are displayed, but also manually set keywords and expanded keywords are shown on the screen. All the words displayed on the screen can be used as an search keyword by dragging and dropping the word into the text area shown in the lower part of the screen.
As stated before, precision of voice-to-text speech and search of detailed search condition is not necessary high. In that case, user may be frustrated if we only display "no results found" or totally irrelevant result. In this research we tried to display as much words as possible on the screen. By viewing those words, it it probable that some words on the screen might be interesting for the user. If so, user can start new search via interaction on the display, not by voice recognition. EVALUATION We have conducted two types of evaluation so far. First, we evaluated how effective search algorithm is for various user input. Since we would like to evaluate effectiveness of our algorithm, we used text input rather than voice input to avoid error caused by speech recognition.
We have chosen fifty sentences from Q&A site about real
estate search[
As a result, 34 out of 50(68%) sentences can be evaluated as success. 9 out of 34 sentences which are evaluated as success include manually defined extend keywords. If we remove them from the result, 25 out of 41 (61%) sentences can be searched successfully using proposed algorithm.
Second, using current system, we conducted simple user evaluation. We have explained system’s concept and function, and put the system beside, we had conversation about what type of real estate property the user is interested in. Five users have participated in the test. During the user test, we got consistent response from the user. Every participants see the importance of system’s assistant role. In some cases, the user can find the interesting results. Four out of five participants see detailed search condition that they have never searched real estate property with before. We also observed that the user quite often try to use the voice interface as an main interface. Even though current system only extracts and search with detailed search condition, the user tried to search with voice sentence like "Search property in Tokyo area". Even though we designed the system so that the user can set those search condition with GUI, the user often forget that.
From the result of user test, we realized that interaction design should be improved so that user does not misunderstand that system can recognize every request that user might have. In the current design, search condition setting GUI is hidden while the system listen the conversation. In this case, the user expects system can understand any word what they say. To avoid such a misunderstanding, we need to show search condition setting GUI upfront. When the user speaks, and do not operate the GUI, we can show current voice recognition interface over the search condition setting GUIs.
CONCLUSION AND FUTURE DIRECTION We have developed the real estate search system which recommends detailed search conditions from users’ conversation. Evaluation result shows the mixed results. We could confirm the potential of proposed algorithm. However, we also recognized that user interface design should be improved. Based on the evaluation results, we will redesign the system interaction, and will conduct further user test to evaluate how effective system can support the search of real estate information.