This page looks at one way to use advanced Internet searches to estimate certain probabilities.

The notes date from the Presidential election of 2000 when Bill Clinton was President and Al Gore was the Democratic nominee and George Bush (son) was the Republican nominee.

At that time, Alta Vista was a popular search engine and Google was an up-and-coming search engine. One of the problems with Internet searches is that there may be too many hits for a given search string. The other problem is that it may be hard to determine how valid the information is that is found with the search.

Many search engines offer advanced search options that support logical operations, resulting in fewer hits that must be looked at.

Many offer shortcuts for advanced logical operations.

Each search engine has its own syntax for advanced queries.

The results that follow use the Alta Vista search engine, at http://www.altavista.com, as Google did not seem to handle the logical or operation correctly (or, there was something that we could not get right).

The results that follow were done in March, 2000, an election year, in which Bill Clinton was President, and Vice President Al Gore was on his way to winning the nomination to run for President.

• Logical Not is "!".
• Logical And is "&".
• Logical Or is "|".
• The plus sign "+" includes pages with the next operand.
• The minus sign "-" excludes pages with the next operand.
• The operator Near is "~", as in "Clinton ~ Gore" would find the text "Clinton" that is within 10 words of the word "Gore".

We will search for Clinton and/or Gore. Remember, though, that we will also find hits that do not refer to President Clinton or Vice President Gore.

This would become more problematic when Bush was elected president as there are many hits for the plant variety of "bush".

• Clinton & Gore would find pages with both Clinton and Gore.
• Gore & ! Clinton would find pages with Gore, but not Clinton.
• Clinton & ! Gore would find pages with Clinton, but not Gore.
• Clinton | Gore would find pages with Clinton or Gore or both.

• Clinton would find pages with Clinton.
• +Clinton would find pages with Clinton.
• +Clinton+Gore would find pages with both Clinton and Gore.
• +Clinton-Gore would find pages with Clinton but not Gore.

Notice that the numbers do not quite add up, but are very close.

Here is some probability for those interested.

P(Clinton | Gore ) = 118,889 / 511,253 ≈ 0.232.
P(Gore | Clinton ) = 118,889 / 1,482,243 ≈ 0.0802.

Thus, it is about 0.232/0.0802, or about 2.9 times more likely that Clinton is found on a page with Gore on it than Gore is to be found on a page with Clinton on it.

Suppose that Gore wins the election and becomes President. How do you expect that these search results would change?