Anonymous service person looking over British Army Training Unit Suffield BATUS battlefield

Frontline Tech: Could Chatbots Be The Answer To Battlefield Questions?

It is not surprising the military would want their own chatbot to help navigate their masses of stored data, writes David Hambling.

Anonymous service person looking over British Army Training Unit Suffield BATUS battlefield

By David Hambling, technology expert

The Ministry of Defence (MOD) is developing its own chatbot to get vital data to soldiers in the field fast.

Voice-activated virtual assistants like Siri and Alexa are now commonplace, and many businesses use chatbots rather than humans to answer questions online.

Instead of having to wait for a human to be available, you can get answers immediately.

Thanks to artificial intelligence, computers can interact in normal language and provide useful information.

It is not surprising then that the military would want their own chatbot to help navigate their masses of stored data.

The MOD’s Defence Science and Technology Laboratory (DSTL) is running the project.

Military requirements mean that a new type of chatbot is needed, and they cannot just purchase commercial software off the shelf.

For one thing, it will have to work over the limited bandwidth provided by a secure tactical radio connection rather than over 4G or Wi-Fi.

For another, it will only access data on the MOD’s satellite network or stored locally rather than using the internet. Military systems are kept separate for obvious security reasons.

ATAK being used on mobile phone by British soldier
British personnel tested battlefield communication technology known as the Dismounted Situational Awareness Tool last year (Picture: Crown Copyright/Cpl Tom Evans).

After a competitive evaluation earlier this year, the MOD selected British geospatial and data company Envitia for a two-year contract in August.

They aim to provide a chatbot with the aim of making things easier than the current awkward way of finding data.

"Currently the user needs to access document stores, such as disparate SharePoint sites or secure file shares to find the information," says Envitia CEO Nabil Lodey.

"This is both time consuming and not necessarily easy to navigate."

That chatbot locates and retrieves information automatically.

It will have a text interface, with the ability to work both online and offline, and will be able to carry out a conversation, asking questions to clarify exactly what the user needs.

The initial version will be text-only, but Mr Lodey says that a voice-activated version may be developed later if needed.

"The chatbot will be able to have a conversation with the user to discover exactly what is needed,” he says.

"If there is ambiguity in the search it is performing in the background, it may ask the user for more context to try and narrow down the search results."

British Army Training Unit Suffield BATUS battlefield
Bandwidth limitations mean that users in the field cannot always download such large files.

As anyone who has searched the internet knows, the problem is sometimes too much information and the number of possible answers you get back.

Another issue is that the detail you want may be buried inside a massive document, and bandwidth limitations mean that users in the field cannot always download such large files.

“Our Chatbot solution will filter out irrelevant material, and then provide the vital information that is needed for mission success,” says Lodey.

Development will be based on a machine learning approach in which the chatbot learns about common problems and requests, and works out the best answers to them.

It builds on previous DSTL work on a Federated Search capability, which can search several different resources at the same time to get answers quickly.

There is no limit to the sort of information which the chatbot could retrieve, and it could be used to access all sort of data depending on what the DSTL decides.

“We would like it to have practical applications for soldiers, covering all kinds of scenarios from vehicle repairs and navigation to military operations and deployment,” says Lodey.

One use in the initial specification is extracting data from the Single Intelligence Environment which brings together data from different sources about “objects of interest.”

These may include people, places or things in the current theatre of operations.

So soldiers may be able to get information about a particular village or a high-ranking insurgent, which may have been gathered from human sources, signal intercepts, satellite sensors – or simply picked up from the internet.

Soldiers may be able to get information which may have been gathered or simply picked up from the internet.

Having a single source should mean that every bit of data is immediately available to the people who need it, rather than being filed away.

This is not quite the first military chatbot.

The US Army launched ‘Sergeant Star’ in 2006 as an interface on their GoArmy recruiting website. Sergeant Star can communicate either through a keyboard interface or verbally, and has answered over seven million questions from potential recruits.

Mainly Sergeant Star helps with queries about what types of job are available, pay and conditions and health and fitness requirements.

But about 10% of the questions have been about Star himself, such as whether he is married, or if he has a dog.

Being a chatbot, Star answers every question, even the ridiculous ones - apparently he is married to Mrs Star and has a rottweiler named Chomp. 

Star is essentially a civilian chatbot with military styling to provide recruitment information to civilians.

By contrast, the MOD chatbot is a new type of development to feed information rapidly and efficiently to soldiers in the field.

If it proves successful, it is likely to be copied by other militaries around the world.