Feature

Anatomy of a Knowledgebot

by Mark Gerow

What is a Knowledgebot?

“Natural language processing” (NLP) in products like Apple Siri, Amazon Echo, and Google Home let you to talk to your car, phone, or digital assistant and have that device understand what you are saying. In addition, it is what allows that device to answer back. Such voice user interfaces (or VUIs) are becoming a prominent, if not the dominant means of efficiently interacting with computers and devices of all sorts.

The term “knowledgebot” refers to an NLP-powered chatbot (“bot”) intended for use by knowledge workers needing access to a particular set of information and processes in order to get their work done. In many ways, knowledgebots are the successor to (or at least a significant complement of) the enterprise intranet. If an intranet is a modern-day bulletin board, where seekers scan the contents looking for what they need, then a knowledgebot is a concierge that will find the information sought and deliver it to them.

What can a knowledgebot do?

A few real-world examples of what a knowledgebot can accomplish are:

• Book a visiting office at another firm location
• Get the current NDA, waiver letter, term sheet template, or other standard template
• Obtain the correct billing rate for an attorney, paralegal, or other timekeeper
• Get a copy of the firm’s charitable giving policy via email
• Find colleagues who have worked on matters for a particular client

All of which can be performed using a mobile device or web browser. Of course, your firm’s intranet can provide all of the above answers and workflows, but a knowledgebot can do it in a fraction of the time!

Benefits of a Knowledgebot

The specific benefits of knowledgebots can be as varied as the environments in which they are used. The beneficiaries of knowledgebots fall into four broad categories: seekers, experts, curators, and the firm.

Seekers are individuals who interact with the bot in order to obtain an answer or complete some action. The benefit to the seeker is straightforward - they get what they need faster, and with less effort.

Experts are those who provide the raw material for the answers and actions the bot provides. Experts benefit because their expertise can be shared more widely, and with less effort.

Curators are responsible for maintaining a robust knowledgebase, and benefit from the knowledge and workflow usage data the knowledgebot provides.

The firm or legal department benefits from the resulting collective productivity gains. It also benefits by having a deeper understanding of the knowledge assets (question/answers and workflows) that its professionals use to do their work. In effect, the knowledgebot platform provides an x-ray image of information and process flows throughout its operations.

Put simply, firms using knowledgebots will improve the collection, curation, and dissemination of knowledge assets significantly.

Speech Recognition

Although not absolutely required, seekers have come to expect that they can speak to a bot and have it understand what they say. When connecting to a bot, the hardware or browser software takes care of the speech recognition.

Intent Recognition

Once speech is converted to text our knowledgebot needs to determine its meaning, or “intent”. A request’s intent consists of several components, including:

• The form of the request
• Any key words or proper nouns that refine the meaning
• The context within which the request is made

For example, the question “what is [colleague’s name] standard billing rate?” would be broken down into:

• “billing rate” - i.e. get me a billing rate
• “standard” - i.e. which type of rate to return, and
• “[colleague’s name]” - the individual who’s rate I want

In the above example, once recognition is complete, the bot can search the firm’s financial system to return the corresponding rate.

Request Processing

Request processing is where the knowledgebot connects to the data and workflows that provide the answers and actions to the seeker. It does so by passing information from the NLP AI to some service (referred to as a “webhook” or “fulfillment service”), which then queries databases, calls other services, or initiates workflows as needed. This service also determines the response, which is sent back to the NLP AI (Alexa, DialogFlow, Cortana, etc.) which in turn responds to the seeker.

Speech Synthesis

In order for the seeker to hear the knowledgebot’s reply, the text response needs to be converted to an audio signal that can be played back through the seeker’s device or browser. This process is called speech synthesis.

Speech synthesis may be handled in hardware on a phone or AI assistant device, in a browser, or via a web service such as Amazon’s Polly which returns an audio file containing the spoken equivalent of a text string. In any case, there are several benefits to returning answers as speech, including:

• It allows hands-free interactions, as when a professional is en route to a client meeting
• IWe humans are “wired” for speech

Conclusion

NLP AI and supporting technologies have advanced to the point where it is both feasible and practical for firms to develop knowledgebots that allow attorneys, staff, and clients to access information and workflows in a fraction of the time possible through intranets and extranets, providing greater efficiencies and mobility for all. A secondary benefit of knowledgebots is a deeper understanding of the information and process flows within the firm, which can inform ongoing improvements in legal operations. ILTA