We offer support for multiple languages and ensure fast integrations to our customers’ existing platforms. We believe in the innovation paradox: innovation is lead by small companies, not large enterprises.

The semantic capabilities of our system has to evolve in a language-independent  engine, where each conversation evolves following patterns to be found in previous data.

We plan to extend the technique we are using today from sentences to the whole conversation, allowing a personalisation of the whole experience, with the engine being able to discriminate between users and guide the conversation because of that.

We are already experimenting pattern recognition analysis of conversations, but at the moment strongly prefer not to disclose the results of our research

Neural Networks (NN) are used for the training of the semantic engine. The dimensionality reduction transposition to the semantic space is done through processing corpora of billions of words (Terabyte), and NN are today the only possible tool for the task.

Conversely, we noted that “purely Neural Network” approach, which analyses only the usually scarce amount of sentences produced by even big corporations (counting far below the billion of words), leads to extremely poor results.

To make a paragon in real life, it is like asking a child to learn Finnish and how to answer to customer service questions after just having listened to the few thousands of sentences in a customer service log –sentences which are most of times relatively similar. Computers are good at analysing huge amount of data (Exabyte), and the size of logs recording past conversations in customer service are usually a factor one million below that.

Human language is a statistical association of patterns of words to thoughts. It has been extensively proven that in human being language is not a necessary skill for symbolic thought. Babies after 8 months can already “think”, still they are far from speaking.

Therefore –yes, we do understand the language because we make statistical associations, exactly like a human beings understand that, most of the time, when they hear the word “dog” a four-legged animal is going to be around.

Semantic (=meaning) extraction in Jenny is done through a dimensionality reduction process and the creation of 300 features for sentence. This is done through Bayesian inference of the probability of words having similar role in the syntactical tree of the sentence.

The resulting 300-dimension space is therefore resulting in a semantic classifier, allowing a virtually infinite number of sémata (meanings, signs) to be detected.

We have developed a software solution that enables companies to build and deploy chatbots onto various messaging channels in their own terms. We bring Artificial Intelligence to any support operations of an organization with an Open Source automated chat solution.

We solve the expensive problem of repetitive conversations in chat support, through chat automation on any language.

We support any language that our customers require. We have built a system that enhances the data owned by companies and we are developing a fully multilanguage platform using unsupervised algorithms for all languages.  Most of our competitors focus solely on English. Our plug-n-play model offers our clients an easy and fast setup, and the system enables customer to enhance the data through day-to-day workflow.

We offer both an on-premise solution as well as a service.

GetJenny has the ability to deliver on-premise cutting-edge technology that integrates directly into customer’s infrastructure. We can easily implement chatbot systems able to read and write to existing databases.

We currently offer integrations with the following channels

  • Elisa Chat yrityksille.elisa.fi/chatpalvelu
  • Facebook www.developers.facebook.com
  • Genesys www.genesys.com
  • Giosg www.giosg.com
  • Kik www.kik.com
  • Lekane www.lekane.com
  • Line www.line.me
  • LivePerson www.liveperson.com
  • Ninchat www.ninchat.com
  • Skype www.docs.microsoft.com/en-us/bot-framework/channel-connect-skypeforbusiness
  • Slack www.api.slack.com/community
  • Telegram www.telegram.com
  • Zendesk www.zendesk.com

We are currently working on multiple integrations, such as:

Open Source means we make the source of the software available to other people.

It does not mean it is free, but it means it is a fully functional software, that is easy to install and use.

If external developers want to use StarChat with a click, they must work on it. We are not going to work on a ready to use UI at the expense of what we feel has priority (e.g. improving the NLP).  It does not mean a perfectly documented software.

We’ll try to document StarChat as good as possible, but again, there are priorities. If developers want to use StarChat, they must be knowledgeable enough to understand how it works.

We want developers contributing to StarChat, not just us contributing to this project for others to use.

Being Open source enables enterprises to mitigate their risks. Every time an enterprise is buying from a startup it is a risk because they don’t know if the startup will be there few years down the line after purchase. We are open source, which means that the software will always be available for further development by anyone, irrespective of the sustainability of our start-up.

In brief, we want to create a community of developers working on StarChat, not to give software for free to everyone. Our development team is too small to create alone a comprehensive multilingual chatbot –therefore we ask the community to collaborate. In return, they’ll have the possibility to use StarChat with the same rights and duties, be able to sell services based on it, and provide the code to the community.

For instance, if a developer thinks that the documentation is not clear enough, s/he should propose changes. If the documentation is good enough for us, there is no reason why we should improve it.

Collaboration is achieved through modularization. This will allow developers to implement “plugins” to add functionalities to the package.

You can join the community through www.github.com/getjenny

We have open sourced StarChat to foster collaboration on our technology

In the 1990s it became clear that software has become too complex to be managed by a single organisation without the help from the community. Successful Open Source software solutions have eclipsed proprietary competitors since then, with rare exception of Google MapReduce that emerged due to peculiar labor market reasons, explosion of dot-com bubble and brightest minds working for Google. Even Google now open sources its technology to keep pace with evolution.

Often, we are asked how we think we can make revenues where our main product, the conversational platform StarChat, is open sourced –i.e. can be copied by other developers.

Briefly: no successful open source project has been “stolen”.


  1. If a company has a whole team of engineers and scientists equal or better than the one developing the open source project, it would make no sense not to start from our quality software. This in turn would make StarChat better due to the collaboration.
  2. If the team is at a lower level of development ability, they do not pose a threat. Jenny as an example, the competitor could use StarChat, but not having the deep know-how necessary, they would be unable to integrate it with the customer’s technology (all our integration software is proprietary)

StarChat, is the core technology, which is a scalable conversational engine for B2B applications. Developers and companies can start with the open source technology very easily. The easiest way is using two docker images. For Jenny to be able to have good matching and statistics, companies might want to index a corpus which is hidden from results. For companies to begin quickly, they can upload their FAQ’s and activate it for Question and Answer chatbot.

The engine uses meaningful words, these can be used as keywords or regular expressions such as age or city. The system is able to memorize the conversation by using Traverse states. The engine stores the information to continue the pattern of the conversation that is depended on the input user has said or information needed to go forward with the pattern.


CPU: quad core Intel(R) Atom(TM) CPU  C2750  @ 2.40GHz

Memory: 4GB

disk: 50GB of ssd disk available for StarChat


Linux, any recent distribution e.g. ubuntu 14.04


The service binds on a tcp port and serves incoming connections through the HTTP protocol

Developed in Scala: a language thought with distribution computing in mind, with easy integration with Apache Spark, IOHO the most promising ML project available now. Semantic similarity is done now with Elasticsearch. We are working on a document similarity “doc2vec” based on word2vec models–starting from Finnish and Italian.

We are an open source company, you can find the source from https://github.com/getjenny

More in the section “open source”

We have published our code under GPLv2.

Contributions are in form of “pull requests” which means a request to merge their modification on our branch, after a review we merge the modification. Each pull request is registered and it is possible to determine who has contributed and in which part of the code.

The license and the terms are also published on the repository and the copyright is ours. The license in the master branch is GPLv3 but we will change to GPLv2.

We are designing a big modification to the state machine mechanism at the moment. The modification will add the support for regular expressions and introduce the means to work with probabilities and scores. This component is critical since it is fundamental to add to the intelligence of the chat. The roadmap will be communicated using github or using a dev blog if necessary.

We are able to go live in 1 month, and our approach enables non-technical people to integrate our conversational agents to their systems. We bring conversational computing to the mass market.

© 2017 GetJenny. All Rights Reserved.