How can product-related research be carried out in the future?
Posted: Wed Feb 19, 2025 6:40 am
Language models (LLMs) learn based on the frequency of co-occurrences, i.e. in the context of e-commerce from co-mentions of attributes with the respective product. (see also my article LLM optimization: Can you influence generative AI outputs? )
Which attributes are important for a product entity is determined by the frequency of the attributes that are requested in prompts and search queries.
As already mentioned, future product research will be more usa cell phone number list interactive and contextual. Prompts can be used to add many more layers of context to queries. Here is an example of a product-related prompt.
The topic of the prompt is jogging shoes or running shoes. Contexts are
Age: middle age
Weight: Overweight (weight to height)
Distance: Middle distance
Size: medium-sized person
Frequency: frequent
At this prompt, the different AI systems give us different product recommendations:
ChatGPT suggests specific running shoe models and translates the context from the prompt into corresponding attribute types:
Durability
Foot Type
Gait
Fit
Comfort
Google's Gemini only suggests running shoe brands at first and translates the prompt into the following attributes:
cushioning
Support
Stability
running style
If you ask Gemini for more details about the shoe models, the following shoe models including pictures are suggested.
The recommendations of the two LLMs are similar. The Brooks Ghost, Asics Kayano, Hoka One Bondi and Saucony Triumph are also recommended by ChatGPT.
However, my tests have shown that this is not always the case and product recommendations can vary. This may be due to the different training data.
Why are these products recommended by the LLMs and not others?
These products seem to be frequently mentioned near attributes that have been translated in the respective LLM.
When optimizing for the Shopping Graph, try to mention the relevant attributes in the data sources mentioned above if possible.
Here we used our custom GPT for text analysis via natural language processing to analyze the manufacturer description of the running shoe brand Asics for the Asics Kayano model series.
Which attributes are important for a product entity is determined by the frequency of the attributes that are requested in prompts and search queries.
As already mentioned, future product research will be more usa cell phone number list interactive and contextual. Prompts can be used to add many more layers of context to queries. Here is an example of a product-related prompt.
The topic of the prompt is jogging shoes or running shoes. Contexts are
Age: middle age
Weight: Overweight (weight to height)
Distance: Middle distance
Size: medium-sized person
Frequency: frequent
At this prompt, the different AI systems give us different product recommendations:
ChatGPT suggests specific running shoe models and translates the context from the prompt into corresponding attribute types:
Durability
Foot Type
Gait
Fit
Comfort
Google's Gemini only suggests running shoe brands at first and translates the prompt into the following attributes:
cushioning
Support
Stability
running style
If you ask Gemini for more details about the shoe models, the following shoe models including pictures are suggested.
The recommendations of the two LLMs are similar. The Brooks Ghost, Asics Kayano, Hoka One Bondi and Saucony Triumph are also recommended by ChatGPT.
However, my tests have shown that this is not always the case and product recommendations can vary. This may be due to the different training data.
Why are these products recommended by the LLMs and not others?
These products seem to be frequently mentioned near attributes that have been translated in the respective LLM.
When optimizing for the Shopping Graph, try to mention the relevant attributes in the data sources mentioned above if possible.
Here we used our custom GPT for text analysis via natural language processing to analyze the manufacturer description of the running shoe brand Asics for the Asics Kayano model series.