Akvelon | Case Study: Social Chat Bot
21460
post-template-default,single,single-post,postid-21460,single-format-standard,ajax_fade,page_not_loaded,,qode-theme-ver-8.0,wpb-js-composer js-comp-ver-4.9.2,vc_responsive
 

06 Jan Case Study: Social Chat Bot

Business Need

Although internal networks can help employees in distributed teams communicate, some still feel disconnected from other teams within the company. The goal of this project was to connect employees from all teams across the globe and encourage them to engage in the company’s social media activities, leading to an increase in employee motivation, fellowship, and collaboration throughout all distributed teams.

Solution

Microsoft Bot Framework helps to leverage our AI model across multiple bot channels for both enterprise and consumer scenarios. Akvelon started with a Skype chatbot channel. This chatbot utilizes natural language (through LUIS) and learns employees’ interests and preferences. With this knowledge, it then suggests relevant news or posts from all of the company’s various social media feeds (Twitter, Facebook, and LinkedIn, etc).

 

To prevent the bot from sending out re-posts and duplicates from multiple social channels, we needed to filter out similar and identical posts from the queue. Akvelon developed AI that analyzes the text of each post’s semantic likeness to combine similar posts from various social media feeds owned by different organizations throughout the company.

 

The text similarity model was developed using a hybrid Corpus and Knowledge-Based similarity method that utilizes a cosine similarity measure. Explicit Semantic Analysis and NLTK library was used for the model implementation. To identify synonyms from different news sources,  we used the WordNet lexical database. Specific domain related keywords are looked up with a trained group of linguistic two-layers NN models (word2vec).

 

Benefits and Results

Since the release of this bot, measurements show an increase in employee engagement with company social media posts, with a 20%-40% increase in interaction (views, likes, comments, shares) across the company’s social media outlets. Employee awareness of each other’s projects and success, resulting in improved employee morale and communication between teams throughout the company.

 

Further development is still in progress. Akvelon  is improving the text similarity algorithm with a recurrent neural networks (RNN) method. Our developers are  also teaching our AI to prepare summaries of posts, articles, and news stories  based on a user’s preferences and previous behavior.

 

Technologies used

AI, neural networks, Microsoft Cognitive Services, LUIS and Text Analytics,  Microsoft Bot Framework, Azure Bot Service, natural language processing, text semantic analysis, Python

Tags:
, , ,


Akvelon

Akvelon

We Make BIG Software Happen. Contact Us