AS THE SEMI-FINAL numbers from the 1980 census roll in, the battle over what they mean -- and what a few key words mean -- gets hotter. The numbers are quite precise, like 635,185, and the words are relatively simple, like "persons" and "enumeration." But their meaning is in dispute becauseconsiderable political power and millions of dollars rest on how they are defined.

Numbers may not lie, but they don't always tell the whole truth either. When the Census Bureau reported last week that the District of Columbia had 635,185 residents, it didn't mean that a total of 635,185 people lived here on census day last spring. It meant it had been able to count that many residents. The number who actually lived here that day is larger because some people, intentionally or accidentally, eluded the census takers.

Although this year's census is regarded as the best the government has ever conducted, the same kind of undercounting occurred all over the country. Because this has happened before, notably in the black areas of big, cities, the statisticians have been studying the subject. They now believe they can estimate with considerable accuracy the percentage of individual racial groups that was not counted. They can therefore produce an estimate of the population of this and other big cities that is likely to be closer to the real number of residents than is the Census Bureau's figure of how many residents it counted.

Which of these numbers should be used to allocate federal aid to the states and apportion the House of Representatives? Spokesmen of big cities, who believe their cities were shortchanged in the 1970 undercount, want the estimates used; spokesmen of rural areas, where counting is easier, want the head count used.

This is where the word "enumeration" comes in. The Constitution speaks of the census in terms of an "actual enumeration." Does that mean, as it appears to, a head count? Or does the other constitutional command -- that seats in the House be apportioned to the states "according to their respective numbers, counting the whole number of persons in each state" -- permit use of an estimate if the estimate is shown to be more accurate? Those are the questions that have got the census results tied up in the courts.

But that's not the end of the problem. There is a strong move in Congress to declare that aliens here illegally are not among the "persons" to be counted for apportionment purposes. The argument for this proposal is more emotional than rational. The same part of the Constitution draws a distinction between "citizens" and "persons" and the latter has long been defined to include aliens, both legal and illegal. In addition, the original apportionment clause included among those to be counted as persons indentured servants and three-fifths of the slaves, neither of whom had as many rights as illegally present aliens now do.

If the big cities keep pushing the estimate-versus-head-count argument and Congress decides to argue about out-of-status aliens, problems of defintion will probably be settled by the courts. That could mean a long delay. It would be better if the arguments were dropped and the Census Bureau's head count accepted, at least for apportionment purposes, this time. Then the issues could be fought out at a more leisurely pace before it is time to count everybody again in 1990. If that doesn't happen, those numbers the Census Bureau will be releasing soon will be interesting -- but nobody will know what they mean in terms of dollars and political power for a long time.