Anúncio

Implementation of FAQ Pages using Chatbot

10 de Jul de 2018
Implementation of FAQ Pages using Chatbot
Implementation of FAQ Pages using Chatbot
Implementation of FAQ Pages using Chatbot
Implementation of FAQ Pages using Chatbot
Anúncio
Implementation of FAQ Pages using Chatbot
Implementation of FAQ Pages using Chatbot
Implementation of FAQ Pages using Chatbot
Implementation of FAQ Pages using Chatbot
Próximos SlideShares
Implementation of FAQ Pages using ChatbotImplementation of FAQ Pages using Chatbot
Carregando em ... 3
1 de 8
Anúncio

Mais conteúdo relacionado

Similar a Implementation of FAQ Pages using Chatbot(20)

Anúncio

Implementation of FAQ Pages using Chatbot

  1. Implementation of FAQ Pages using Chatbot Sarath Nair ,Student,Dept. of CSE,BMSIT&M,Bangalore, saruichu@yahoo.in Sriram AD ,Student,Dept. of CSE,BMSIT&M,Bangalore,sriram_ad@yahoo.in Sagar SP ,Student,Dept. of CSE,BMSIT&M,Bangalore,spsagar9181@gmail.com Tushar Sinha,Student,Dept. of CSE,BMSIT&M,Bangalore, tushar1608@gmail.com ABSTRACT A website or an app is a customary way that a business adopts to provide their services to their customer base. However, given the limited storage in the phones, not many users will be willing to download an app to get their queries addressed. Going to company website is also time consuming. In general, when the consumers face issues, they reach out to the customer support. More often than not, it takes a long waiting time to reach the customer service representative. Not to mention, these calls are not always satisfactory. Interacting with customers and retaining those customers becomes difficult for the businesses with a wide audience to cater. Chatbots provide an option that can be used by businesses to address the general queries of the user. These are chat-based software that understand anything user types or says and accordingly replies and takes actions. The recent developments in the field of artificial intelligence have made chatbots more intelligent and adaptable for being a substitute to FAQ pages. Keywords—Chatbots, Artificial Intelligence, FAQ pages INTRODUCTION FAQ page is section of a website that becomes a go-to destination for the customers. This page is one of the most important pages on the website and helps the business increase its online presence and drive qualified traffic to its site. Many companies have realized this and have spent a lot of time and effort in improving the content of their FAQ pages and also the way this content is being served to the user base. Chatbots offer a simple solution to this issue. International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 187 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  2. Chatbots have been there for some time now. They have been gaining a lot of popularity ever since they hit the market. Despite this not a lot of companies have used them as a replacement for their FAQ pages. The reason has something to do with the way chatbots have been implemented in the past. The chatbots of the past have not been intelligent. These chatbots have not showcased an ability to have human like conversation with the users. The traditional chatbots have used buttons and decision tree models, a technique that has been employed by call centers in past with limited success. In technical terms, these chatbots were not conversational. There are lots of problems involved when users are forced to choose from a set of options. It is likely that at some point of time during the conversation, the option they need is not part of that set. Moreover, there is a chance that they try to ask question to a user in a form that is not understood by the chatbot. Another issue worth considering is that if the user changes his mind during a conversation, he won’t be able to go back on that decision and will be left with no options but to start over. The industry needs a conversational chatbot now more than ever. In this paper, we aim to propose a model of a conversational chatbot that can be used by the industry as a substitution for their dull FAQ pages. Deep learning concepts can be used to create an intent and entity recognition model. Intents are classes that highlight the main communicative essence in the user input. Entities are words of value in a user input. Each user input may contain zero or more entities contain key information important to the ongoing conversation. Common examples of entities include names of organizations, locations and prices. Once the intent and the entities are recognized, the remaining part of the job is to generate a suitable response from the knowledge base. Just replying in text format won’t be enough, to reply more human like experience, it must provide voice interaction as well. RELATED WORKS Most of the chatbots in the industry are rule based. The knowledge of these chatbots are stored in the form of patterns and templates. When the user query matches one of the patterns, the response stored in the <template> is sent to the user. The <pattern> could be a simple sentence like “what is income tax?” or a string with regular expression like “what is * ?” .The <pattern> and <template> are handwritten.[1] An inherent problem with this approach is finding an appropriate algorithm to match user queries to a particular <pattern>. Eliza, one of the earliest chatbots created in 1964 at at the MIT Artificial Intelligence Laboratory by Joseph Weizenbaum used an interesting approach. International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 188 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  3. Eliza used incremental parsing for pattern matching. All the words in the user input were parsed and looked for in the dictionary. Each word was given a priority based on importance and stored on a keyword stack. The word on the top of the keyword stack was tried to be associated with one of the patterns. Based on the pattern match, a suitable reply stored in template was generated. In case an association with a pattern could not be made, a default reply like “I see”, “Please go on” was displayed. [2] ALICE, a later implementation of chatbot, inspired by ELIZA used a different technique for pattern matching. The knowledge of ALICE was stored in Graphmaster, a graph with nodes and edges. The path to every leaf node is a sentence or a user query, the reply for which is stored in that particular leaf node. The problem with all the rule based approaches is that the rules must be provided by the programmer .A lot of time and effort is required in writing such rules. [3] PROPOSED WORK A lot of developer’s time will be saved if they can utilize the large sets of chat logs on various chat platforms like Twitter and Facebook. Instead of classifying the sets into questions and responses, deep learning techniques can be used to recognize intents and entities in the user input and map these intents and entities to a suitable response stored in a database. As shown in Figure 1, a user starts the conversation by asking a question which is sent to the ChatBot. The ChatBot then processes the input query and generates the response for the user which is sent as a reply to the user. Fig 1: Block diagram of intent and entity recognition process International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 189 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  4. If the ChatBot is unable to find suitable answer or response for the input query it sends the query to the admin. The admin can check the log file to find the unanswered queries and adds the suitable response in the form templates and updates it back to the ChatBot. This continuous process helps in learning through user interaction and if the same question is asked again the ChatBot is able to answer it. ChatBot controls conversation flow based on the context of the user’s requests and responds with natural language phrases to provide direct answers, requests additional information or recommended actions that can be taken. Fig 2: Basic Conversation Flow Figure 2 provides a high level description of how a chat client could be used to leverage natural language processing to assist with access to content or perform data queries. PROPOSED WORK A lot of developer’s time will be saved if they can utilize the large sets of chat logs on various chat platforms like Twitter and Facebook. Instead of classifying the sets into questions and responses, deep learning techniques can be used to recognize intents and entities in the user input and map these intents and entities to a suitable response stored in a database. As shown in Figure 1, a user starts the conversation by asking a question which is sent to the ChatBot. The ChatBot then processes the input query and generates the response for the user which is sent as a reply to the user. If the ChatBot is unable to find suitable answer or response for the input query it sends the query to the admin. The admin can check the log file to find the unanswered queries and adds the suitable response in the form templates and updates it back to the ChatBot. This continuous process helps in learning through user interaction and if the same question is asked again the ChatBot is able to answer it. International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 190 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  5. Fig 1: Block diagram of intent and entity recognition process ChatBot controls conversation flow based on the context of the user’s requests and responds with natural language phrases to provide direct answers, requests additional information or recommended actions that can be taken. Fig 2: Basic Conversation Flow Figure 2 provides a high level description of how a chat client could be used to leverage natural language processing to assist with access to content or perform data queries. RESULT Figure 3a: RASA Server International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 191 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  6. Figure 3b: Flask Server Figure 3a and 3b show the RASA server and Flask server. The Flask server takes user input from client interface and passes it onto the RASA server. Figure 4: RASA server’s response Figure 4 shows the RASA server’s response to a query in json format. The figure shows how intents are ordered in terms of the confidence order. International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 192 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  7. Figure 5a: Interactive chat client interface Figure 5b: Interactive chat client interface Figure 5a and 5b show the interactive chat client interface. It shows the send button, speak button and listen button. These buttons are used for textual and voice interaction with the Chatbot. International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 193 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
  8. CONCLUSION By enabling textual and voice interaction between Chatbot and the user, the overall user experience is enhanced. Chatbots have shown the potential to replace the website based implementation of FAQ pages. It is a technology that can allow users to have natural conversations to access content and services. Chatbot typically take the form of a chat client, leveraging natural language processing to conduct a conversation with the user. References [1] Wallace, R. S. The Anatomy of A.L.I.C.E. Retrieved from http://www.alicebot.org/anatomy.html [2] Lokman, A. (2010). One-Match and All-Match Categories for Keywords Matching in Chatbot. American Journal of Applied Sciences, 7(10), 1406– 1411. http://doi.org/10.3844/ajassp.2010.1406.1411 [3] AbuShawar, B., & Atwell, E. ALICE : Trials and Outputs. Retrieved from http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid=S1405- 55462015000400625 International Journal of Computer Science and Information Security (IJCSIS), Vol. 16, No. 6, June 2018 194 https://sites.google.com/site/ijcsis/ ISSN 1947-5500
Anúncio