top of page

Optimizing Trucker Capacity with NLP

Updated: Mar 4

One of a startup we've worked with is having problems obtaining the availability of trucks for freight needs. What they have is a way to match freight shipping needs to trucks thus getting the available capacity to service their freight is the main problem for shippers.

For some context, shippers are either manufacturers, distributors, or even brokers with freight that needs to haul across USA. Truckers are either owner-operated tuckers or companies with multiple trucks that needed freight shipment to move.

The Problem:

To get more data from truckers into the client's system, they have to build a simple UI form, which has a way to input individual trucks into the system daily for freight matching. The reason is that every day, trucks are moving, and location changes every hour, so we need to find an optimal freight pickup point that matches based on equipment, direction, and price, however, no single customer of our client used the UI form at all. There are varying needs and customize fields that don't match what truckers have in their system. There is also a lot of complexity in determining routes, including different underlying systems to fill the form. And lastly, Data encoders on the truckers' side don't like adding information manually every day to our client's system, since this is one part of their system to be dealt with, they also need to arrange for billing, collections, driver scheduling, dispatching, operations and all, the system is ever is old and outdated. Technology is also new, they are used to Fax, EDI, and SMS. Changing their behavior is a tall order, and filling in information is a big lift.

What Truckers do in the meantime, is send its capacity to our client via Email, and most of the truckers do that to their contact brokers and/or Shippers. So it is a common practice to do that.


We employ the obvious, use NLP, and extract all the important records and/or information from the email and inject that information into the client's system, so we can continuously infer, and match.

So the flow of data is like this

Broker/Freight —> Loads —> Triseed Client —> Truckers

Truckers —> Truck Capacity —> Triseed Client <— brokers/Freight

One bright spot here is that trucking companies provide their capacity to brokers and freight companies using email, however, since it is freeform, each trucker provides data the way it is convenient for them, even attaching pdf or csv.

To obtain more truck capacity we need to use a different way of capturing the data we've recommended and used Machine Learning to retrieve those data.

Email 1

Email 2

Email 3

Here are the challenges of free-form email.

Dates: Relative dates, Tomorrow, Today, Date Format (ISO), mm/dd/yyyy or mm/dd or dd/mm, etc.

Location: Source, Destination, LAX, MCO, Airport Codes, Zones and Region, Midwest, Northeast, Z1, etc.

Equipment Type: Boom, 42ft, flatbed, container van, etc.

We added a classifier and transformation layer, to further transform the data, to the ones that the client's system exists, We have to make sure the data sent is what the client's system expects so we built some translation, i.e. LAX is really Los Angeles. Midwest is composed of (IL, IN, IA, etc.. )

Also, there are a few things that are not parsable, and trucks capacity is not completely extracted, however, the precision and recall results are high for the data we've trained so we are happy with the results and able to augment data we don't support based on correct data matching.

Learn how we solve this for our client using NLP, Streamsets, Python, and among others. Please contact us today and let's embark on a journey of innovation and growth together. Reach out to our team at to discuss how we can tailor our solutions to meet your unique needs. Don't wait, take the first step towards transforming your business with Triseed!

23 views0 comments


bottom of page