Scientia Professor Boualem Benatallah received his PhD degree in computer science from Grenoble University, France. Today he heads-up the Service-oriented Computing Research Group at the School of Computer Science and Engineering in the areas of web service engineering, business process management, crowdsourcing, data curation, cognitive services, cloud computing and applications.

One of his team’s broad ambitions is to automate the integration of natural language conversations with digital services and processes, through cognitive services and APIs.

Prof. Benatallah says that a core challenge is linking high level and potentially ambiguous user intents to API calls and dynamically synthesised processes. “Put simply we need the technology to comprehend what a person is expressing and the intent of their message.”

A preliminary form of cognitive service technology is a field that most of us use daily and take for granted. Some everyday examples include asking Alexa ‘what’s tomorrow’s weather?’, using a chatbot on a website, or composing an email where the tool drafts a subject line.

To understand the complexities of what’s going on with this technology behind-the-scenes, let’s take the example of Alexa. For Alexa to answer “what’s the weather going to be like in Sydney tomorrow?” it needs to understand what’s been said, recognise the intent of the question and what the user wants to accomplish (i.e. a weather forecast). From there, Alexa has to extract relevant ‘intent slots’, inputting parameter values from the question (i.e. the location being ‘Sydney’ and the date of the forecast being ‘tomorrow’). It then triggers commands that process the user’s intent and ‘call’ the weather API to gather the information and then provides the answer.

Cognitive computing(1) is generally referred to as the simulation of human thought processes    through hardware or software. It focuses on problems and problem statements characterised by ambiguity and uncertainty, and makes these problems computable.

cognitive service leverages cognitive computing techniques to solve such problems by mapping requests or tasks, expressed in natural language, into calls to Application Programming Interfaces (APIs).

An API is a computing interface that defines interactions between multiple software intermediaries. It determines the kinds of requests that can be made, how to make them, the data formats that should be used, and the conventions to follow.

Scientia Professor Boualem Benatallah

Cognitive services technology in law enforcement investigations

Besides cognitive services technology making our lives easier, it is also playing an important role in law enforcement investigations.

One of Prof. Benatallah’s most recent projects(2) researched the support of investigators with the task of collecting and analysing large amounts of evidence items. As information accumulates during an investigation, it becomes vital to keep track of relevant events and detecting possible offences from raw evidence logs. For instance, it can take many collection and analysis hours to sift through thousands of emails and messages to place an entity at the scene of a crime, but now this task can be undertaken much more efficiently thanks to automated curation and tagging.

“We built upon advances in natural language processingword embeddings and knowledge-based enrichment which extracts important information from knowledge graphs, to build a cognitive investigation case management system, called Case Walls,”(3) he explains.

“This enables the technology to recognise entities such as people, events and offences and pull out the useful pieces of information from evidence items.”

In addition to increased efficiency and effectiveness, it also reduces the possibility of evidence being missed by human error.

Making researchers more time-efficient

The research for law enforcement investigations also led Prof. Benatallah on a path to see if they could assist with academic research investigations.

“Students and researchers spend weeks, if not months, reviewing research studies to determine which ones are relevant for their research.

“In collaboration with PhD students and colleagues, we are also investigating support that cognitive services technology can provide to partly automate some laborious and time-consuming research tasks such as research papers screening and coding in systematic literature reviews. This will take out the manual effort of poring through hundreds of research studies to work out which ones are relevant to their research questions and topics.”

Prof. Benatallah says the team is also looking at leveraging this technology in online learning environments.

“In a large online discussion forum, a handful of students may be asking where to find some information, for instance. We are looking to automate this process so that the link to the information is automatically shared, rather than students waiting for the tutor to respond.”

Another significant and challenging issue we are tackling in this area is tagging of vulnerability discovery reports in bug bounty programs.  The objective is to help security professionals to analyse, validate and filter troves of submitted vulnerability reports.

Helping the blind and visually impaired

More recently, Prof. Benatallah started working alongside colleagues in France and Italy prototyping a cognitive services platform for individuals that are visually impaired or blind so that they can communicate with online services using their voice.

“The project is still in the early stages, but we believe it would make a huge difference to people with this disability, giving them access to peer and carer support, work and learning opportunities and so on.

“This would significantly improve their lives and independence.”

If you would like to learn more about Prof. Benatallah’s work in the integration of conversational AI with digital services or discuss potential collaboration opportunities, please reach out via b.benatallah@unsw.edu.au.

(1)S. K. Esser et al.: Cognitive computing systems: Algorithms and applications for networks of neurosynaptic cores. IJCNN 2013, pp. 1-10.

(2)A project funded by the Data 2 Decisions CRC

(3)A project funded by the Data 2 Decisions CRC