How we use word embeddings and AI: from seating arrangements to data analysis

How we solved the seating arrangement: optimisation with smart algorithms

In order to compare different seating arrangements, a way to quantify how good a seating arrangement is required. We achieved this by establishing certain criteria, all of which are associated with a cost if not followed:

• A woman should preferably have at least one female neighbour, and vice versa for men
• Spread based on companies
• Spread based on seating location
• Sit near others with similar interests
• Sit near others of a similar age

In this way, we could derive an actual figure for how good the seating arrangement was.

To optimise the placement, we viewed it as a classic optimisation problem: How do we place 450 participants to minimise the total cost? The algorithm we designed to solve the problem was relatively simple, yet very powerful. It is based on algorithmic optimisation that can be applied to a range of other business problems. By using an algorithm, we could quickly identify the best possible combinations to minimise costs and maximise guest experiences.

Step-by-step: how we optimised the seating arrangement with our algorithm

1. Start with a random seating arrangement.
2. Iterate to convergence/max number of iterations: Select two random people and swap their places if the cost decreases from the swap, see figure below.
3. Swap the places of up to 4 different random pairs of people. This will almost always temporarily increase the cost but is done to escape local minima in an attempt to get closer to our global minimum of the cost function.
4. Iterate to convergence/max number of iterations: Select two random people and swap their places if the cost decreases from the swap.
5. Check if steps 3-4 led to a local minimum associated with a lower cost.
   – If yes, continue from this seating arrangement.
   – If no, revert to the seating arrangement from before step 3 was conducted.
6. Repeat steps 3-5 until convergence.

Embeddings and language models: how we matched interests among participants

To compare participants' interests, as expressed in free text, we used a technique called Natural Language Processing (NLP) and word embeddings. This allows AI to understand and interpret the underlying meaning of words, which goes far beyond simple word matching. For example, compare "I like hanging out with friends" with "Socializing with friends" – even though they contain different words, AI can analyze the common meaning through word embeddings and NLP and match participants with similar interests.

Word embeddings represent words as vectors in a high-dimensional space and are used to analyse semantic relationships. This is particularly valuable for businesses, as AI can interpret large amounts of text data to derive insights from customer feedback, improve HR processes, and even understand trends in market analyses. By understanding the semantic relationships between words, word embeddings can provide companies with a more in-depth understanding of their data. Each word is given coordinates in thousands of dimensions that algorithms have learned by analysing large amounts of text. Words are placed close to each other in the high-dimensional space if they are used in similar contexts.

For instance, even though the words "king" and "queen" do not resemble each other linguistically, they both describe royal roles and have similar relationships to words like "throne" and "crown." Therefore, these words will be "close" to each other in the high-dimensional space.

How our optimised seating plan created engagement and meaningful conversations

So how did it go after we established our framework for an "optimal" seating plan? The response from the conference attendees was overwhelmingly positive. Despite the guests sitting among people from different companies, cities, and ages, they had a lot in common to talk about – thanks to AI helping us match them based on interests.

Here are some comments from attendees:

"Well done on the seating arrangement! It contributed to making it the best corporate dinner I've been to in several years!"

"Fantastically thought out! I found myself next to colleagues I had hardly met before, but we had an incredible amount in common. A very successful setup that truly sparked interesting discussions throughout the evening!"

"The seating arrangement was perfect! Even though we came from different parts of the corporation and various departments, it felt like we shared the same interests, and it turned out to be a really fun evening!"

This shows how the power of data, optimisation, NLP, and AI can solve complex problems – in this case, arranging an almost perfect seating plan for 450 attendees. And the technology is far from being limited to party planning.

If your company is facing complex decisions or needs to analyse large amounts of customer feedback, AI-driven decision support might be the solution. With the help of techniques such as automated text analysis and word embeddings, AI can assist in processing data quickly and providing valuable insights that enhance decision-making.