Map each pronoun to the nearest earlier noun

matchD = {}
foreach p in partOfSpeech["Pronoun"] {
    matched = -1
    foreach n in partOfSpeech["Noun"] {
        if (n < p) {
            matched = n
        }
        else {
            exitloop
        }
    }
    matchD[p] = matched
}

For each pronoun p, reset matched to −1. Walk the sorted noun positions. Every n<p is a candidate; updating matched as n increases leaves the largest position below p. At the first n≥p, exit only the INNER loop and store matchD[p]. Then continue with the next pronoun.

This is the lecture’s deliberately simple positional heuristic. It can choose a grammatically wrong antecedent: “He” at 11 is mapped to “eyes.” at 10. It demonstrates dictionary/list processing, not a complete English pronoun resolver. The −1 result is a sentinel, never a valid word lookup.