Item Content Recommender - Recommend from Interactions

[dogahe]

I have trained a model as follows:

relevant_data = data.select_columns(['full_url', 'text_content'])
model = tc.recommender.item_content_recommender.create(relevant_data, "full_url")

I know I can use the model to find items that have similar text_content by passing their URL as follows:

recommendations = model.recommend_from_interactions(['full_url_1'], 20)

But what I am also interested in doing is to get recommendations for a new item by passing its text_content, something like this:

sf = tc.SFrame({'full_url': ['non/existing/url'], 'text_content':['some text content']})
recommendations = model.recommend_from_interactions(sf, 20)

But the recommendations that I get have nothing to do with the text_content that I have passed in, in fact I get the same result as just doing

sf = tc.SFrame({'full_url': ['non/existing/url']})

Can someone please help?

[vijayirlapati]

Hi, Any solution for this issue? I am also looking for a similar option to get recommendations using item attributes. Also, can we use get recommendations from the item content recommender model using just the item attribute? In the above issue using just 'text content'. Basically i want to get similar items using 'text content'